E. Nemitz et al. Agricultural and Forest Meteorology 105 2000 427–445 429
2. Method and theory
2.1. Measurement techniques The concentration gradients of NH
3
were continu- ously monitored over a 3 weeks period by two contin-
uous ‘AMANDA’ denuder systems ECN Petten, NL; Wyers et al., 1993, each with inlets at three heights.
The concentrations of HNO
3
, HONO and HCl were continuously measured by individual automated an-
nular wet batch denuders ECN, Petten, NL; Keuken et al., 1988, also at three heights. These systems com-
prise the same rotating annular denuder inlets as the AMANDA system, but absorption is provided by ca.
15 ml of 1 mM K
2
CO
3
solution. This solution remains in the denuder for a 1 h cycle before it is pumped into a
test tube for subsequent laboratory analysis. The sam- ples were analysed at ECN by suppression anion chro-
matography using a Dionex AS12A column 4 mm. Given financial constraints, the analysis was limited
to a 5-day period 15–21 June 1995.
Ammonium aerosol gradients NH
4 +
were mea- sured continuously, employing for the first time in
the field the recently developed ‘Steam Jet Aerosol Collector’ ‘SJAC’, ECN, Petten, NL; Khlystov et al.,
1995 with an improved detection limit of 0.05 mm
− 3
. The air stream passing through a denuder inlet of the
type used for the AMANDA is mixed with steam that causes NH
4 +
containing hygroscopic aerosol parti- cles to grow to a size at which they are impacted in
a cyclone. The collected solution is analysed by the same conductivity technique used for the AMANDA
Wyers et al., 1993. In the AMANDA system the sam- ples from the different heights are taken at the same
time and stored in the liquid phase in delay loops for subsequent analysis. In this way errors due to sequen-
tial sampling are eliminated. In contrast, the use of only one SJAC system made it necessary to alternately
sample the air from the denuder inlets at two heights, and sequential sampling errors caused by temporal
concentration changes had to be corrected by linear interpolation Sutton et al., 1993b.
A high volume seven-stage cascade impactor Model 230 with inlet 235, Andersen Samplers, At-
lanta, USA was run for 18 periods of 5–25 h to quantify the size distribution of particulate NH
4 +
, NO
3 −
, Cl
−
and SO
4 2−
in the size range 0.25–6.2 mm, plus backup filter. Ammonia, acid and particle
concentration measurements were supported by profile measurements with up to 10 three-stage filter-packs
Ø = 90 mm, containing a 1 mm PTFE particle filter Micro Filtration Systems, US as well as a NaF im-
pregnated Whatman 41 and a H
3
PO
4
impregnated Whatman 42 paper filter Allen et al., 1989. These
filter-packs were operated for 29 2-h runs at flow rates of about 12 1 min
− 1
, both above and within the oilseed rape canopy. Gaseous atmospheric Cl com-
pounds measured with filter-packs and denuders sys- tems are usually expected to represent HCl, although
Allen et al. 1989 noted some positive interference with low efficiency collection of methyl chloride
CH
3
Cl and methyl tetrachloride CCl
4
, which is here re-examined in the light of emission gradients of
gaseous Cl compounds. The filters of the impactor and filter-packs were
analysed for NH
4 +
at CEH Edinburgh, with a flow in- jection analysis system ‘AMFIA’, ECN Petten, NL,
employing the same type of conductivity detector as the AMANDA analyser. Analysis for NO
3 −
, NO
2 −
, Cl
−
and SO
4 2−
was carried out on a HPLC anion chromatography at CEH Merlewood Dionex DX-100
system with a Dionex Ionpac AS4A 10-32, 4 mm × 250 mm column and a self-regenerating suppressor
ASAR-1. Low concentrations of the filter-pack ex- tracts necessitated an improvement in the standard
HPLC system by the use of a concentrator column Dionex AS4A 10-32, 4 mm × 50 mm. The measure-
ments of micrometeorological parameters were out- lined by Sutton et al. 2000b, while within-canopy
turbulence was measured as described by Nemitz et al. 2000a.
The concentration of NH
4 +
as continuously mea- sured by the SJAC during the period 8–23 June as
well as the results of the impactor and filter-pack runs are presented in Fig. 1. The NH
4 +
aerosol con- centrations showed a clear diurnal pattern with the
concentrations being largest during the day and de- creasing at night-time. The average NH
4 +
concentra- tion measured with the SJAC system was 23 neq m
− 3
0.42 mg m
− 3
. The agreement between the three mea- suring techniques was relatively poor. In particular, the
filter-packs underestimated the concentrations relative to the other methods. Evaporation of NH
4
NO
3
and NH
4
Cl from the particle filter cannot be ruled out, but some uncertainty might also be induced by the use of
independent techniques of the NH
4 +
analysis and by
430 E. Nemitz et al. Agricultural and Forest Meteorology 105 2000 427–445
Fig. 1. NH
4 +
concentration interpolated to 1 m as measured by SJAC, filter-packs and cascade impactor during the period 8–23 June 1995.
the need to correct the data of the SJAC for a reduced capture efficiency of 80.
2.2. Calculation of surface exchange fluxes Concentrations χ of the gases and NH
4 +
aerosol were measured at 1.56 ±0.11 m, 1.95 not NH
4 +
m and 3.28 ±0.04 m within the 1.38 m tall oilseed rape
canopy. From the concentration gradients fluxes F
x
were calculated using the AGM outlined by Sutton et al. 2000b
F
χ
= − u
∗
k ∂χ
∂{lnz − d − ψ
H
[z − dL]} 1
Here u
∗
is the friction velocity, k the von Kármán con- stant 0.41, and ψ
H
the integrated stability function which changes with height z above the zero-plane
displacement d = 1.11 m and atmospheric stabil- ity, parameterized by the Monin–Obukhov length L.
From the measured flux and the concentration at a ref- erence height e.g. z − d = 1 m, the concentration of
the tracer at the mean height of the canopy exchange z
′
may be derived according to χ z
′
= χ 1 m + F
χ
[R
a
1 m + R
b
] 2
where R
a
and R
b
are the aerodynamic and bound- ary layer resistance, respectively Garland, 1977.
Similarly, gradients in temperature T, water vapour pressure e
w
and relative humidity h may be extrapolated to the canopy from measurements of the
sensible and latent heat fluxes. 2.3. Application of the inverse Lagrangian technique
for calculating sourcesink distributions within plant canopies
The inverse
Lagrangian technique
ILT by
Raupach 1989 derives the sourcesink distributions of a scalar in plant canopies from in-canopy measure-
ments of concentration profiles and the turbulence structure. Nemitz et al. 2000a applied the ILT to
derive sources and sinks of NH
3
within the oilseed rape canopy at North Berwick, and the results are ex-
tended here to consideration of within-canopy anion concentrations.
2.4. Estimation of the effect of gas–particle interactions on ammonia flux measurements
In certain situations the effect of GPIC on gradi- ent measurements of NH
3
fluxes may be estimated by analysis of unexpected apparent fluxes of atmospheric
acids or aerosols. Highly soluble gases such as HNO
3
and HCl are generally expected to deposit at the max- imum rate permitted by turbulence V
max
= R
a
+ R
b −
1
, an assumption that was supported by various field measurements Huebert and Robert, 1985; Dol-
lard et al., 1987; Harrison et al., 1989; Müller et al., 1993. Divergence from V
max
has been attributed to
E. Nemitz et al. Agricultural and Forest Meteorology 105 2000 427–445 431
measurement artefacts caused by GPIC within the air Huebert et al., 1988; Sutton et al., 1993a; Zhang et al.,
1995 or chemical interactions on the leaf surface Neftel et al., 1996. Aerosols are expected to deposit
at much slower rates, and deviations such as appar- ent emission of aerosols e.g. Gallagher et al., 1997a
can be used, together with numerical models, to quan- tify the effect of GPIC e.g. Nemitz et al., 1996. Un-
fortunately, during the North Berwick field campaign, background concentrations of atmospheric acids and
aerosols were too small to measure all concentration gradients with the accuracy required for reliable flux
estimates. Other authors have suggested that the ef- fect of GPIC on NH
3
fluxes may be assessed by the log-linearity of the NH
3
profiles Harrison et al., 1989; Yamulki et al., 1996. Looking at modelled concen-
tration profiles, however, it becomes apparent that the curvature of the profiles is not distinguishable from the
scatter of typical measurement data, even if the flux difference between measurement height and surface is
as large as 40, e.g. Van Oss et al., 1998.
In the present case the effects of GPIC on the mea- sured gradients cannot be derived from the compari-
son of measured with expected fluxes, and the inves- tigations are limited to the potential for chemical in-
teractions. Since chemical interactions are driven by deviations from equilibrium, it is useful to assess the
degree of dis-equilibrium between the species and to estimate the kinetic constraint on relaxation. In this
study three different approaches are applied.
2.5. Comparison of measured vapour phase concentration products with the values predicted by
thermodynamic equilibrium considerations
The measured concentration products of NH
3
with the acids K
m
, i.e. [NH
3
]×[HNO
3
] and [NH
3
]×[HCl] can be compared with the theoretical value of the par-
tial pressure product in equilibrium with the aerosol phase K
e
. Potential for aerosol formation would be expected for K
m
K
e
1 and for aerosol evapora- tion for K
m
K
e
1, which would result in an over- and an under-estimation of the NH
3
surface deposition by gradient techniques, respectively Van Oss et al.,
1998. This analysis can provide a measure for the state of the dis-equilibrium and an indication for ki-
netic constraints upon its attainment. K
e
increases with increasing T and decreasing h. For aqueous aerosol K
e
is further reduced by the co-existence of other ions, especially SO
4 2−
, and therefore becomes a function of aerosol composition, which generally varies with
particle size and type Stelson and Seinfeld, 1982. For particles with a radius R
p
0.05 mm, K
e
is increasingly elevated by the Kelvin effect. In the ab-
sence of detailed measurements of the aerosol com- position, here formulations of K
e
were used that have been derived for pure NH
4
Cl and NH
4
NO
3
Pio and Harrison, 1987; Mozurkewich, 1993, and may there-
fore constitute an overestimate. 2.6. Calculation of the chemical time-scale of the
reaction Wexler and Seinfeld 1992 derived a formula which
relates the characteristic time τ
∞
of the achievement of the equilibrium to the size distribution of the NH
4 +
aerosol τ
− 1
∞
= 3D
Z
∞
mR
p
dR
p
1 + λαR
p
R
2 p
ρ
p
3 where D is the molecular diffusivity of HNO
3
0.118× 10
− 5
m
2
s
− 1
, mR
p
the mass distribution function, λ the mean diffusion path length in air 0.065 mm and
ρ
p
the particle density. The most uncertain constant in Eq. 3 is the accommodation coefficient α that is as-
sumed to be in the range 0.001–1 and was set to 0.1 by Wexler and Seinfeld 1990, 1992. Chemical conver-
sion processes are assumed to affect the flux measure- ment if the chemical time-scale τ
∞
is shorter or of similar magnitude as the turbulent diffusive time-scale
τ
d
Kramm and Dlugi, 1994. τ
d
may be calculated as τ
d
= kz1.75u
∗
, where k is the von Kármán con- stant, z the height and u
∗
the friction velocity Brost et al., 1988. The application of Eq. 3 with α = 1
yields a lower limit for τ
∞
, and the effect of chemi- cal reactions is small for values of u
∗
below a critical value u
∗ crit
, at which τ
d
= 0.1τ
∞
α = 1 u
∗ crit
= 10kz
1.75τ
∞
α = 1 4
2.7. Calculation of the coincidence factor of the size distribution of NH
4 +
with those of NO
3 −
and SO
4 2−
aerosols As an indicator for the attainment of chemical
equilibrium, Wexler and Seinfeld 1992 suggested a
432 E. Nemitz et al. Agricultural and Forest Meteorology 105 2000 427–445
coincidence factor C
i
between NH
4 +
and an an- ionic species i as a measure of the degree to which
the two species are found in the same size range
C
i
= 1 − 0.5
Z
∞
m
NH
4 +
R
p
m
tot NH
4 +
− m
i
R
p
m
tot i
dR
p
5 mR
p
denotes the mass distribution function and m
tot
the total mass of the species NH
4 +
or i. Values of C
i
range from 1 total coincidence down to 0 no co- incidence. Since equilibrium theoretically only holds
if the NH
4 +
is present as nitrate and sulphate, the au- thors suggest the restriction of the integration to size
classes in which the chloride equivalent concentration is small 0.1 [NH
4 +
] compared with the ammo- nium equivalent concentration.
3. Results
