class="bi x0 y0 w0 h1"
vii
FOREWORD
The world's terrestrial and aquatic ecosystems are important sources of a number of
greenhouse gases and aerosols which cause atmospheric pollution and disturb the energy
balance of the Earth-atmosphere system. In recent decades the measurement techniques and
instrumentation for quantifying gas fluxes have been improved considerably. Yet, the
uncertainties in the regional and global budgets for a number of atmospheric compounds have
not been reduced due to the large spatial heterogeneity and temporal variability of the factors
that control gaseous fluxes in ecosystems.
Techniques used for extrapolating measurements or properties and constraining results
between different temporal and spatial scales are nowadays referred to as "scaling". All
scaling methods are embedded in the data. Apart from uncertainties associated with the data
used, errors may be caused by generalization of the basic data (e.g. in soil maps, ocean maps).
Moreover, much of the spatial and temporal variation at a detailed level is obscured as a result
of aggregation. Possible errors caused by the use of aggregated or generalized data in models
are generally not explicitly analyzed.
An important step in scaling of gas exchanges between ecosystems and the atmosphere is
the delineation of functional types where distinct differences in structure, composition or
properties of landscapes or water bodies coincide with functions or processes relevant for gas
fluxes. Delineation reduces the variability of state variables, and therefore functional types
form a good basis for measurement strategies and model development.
Models are widely used tools in bottom-up scaling approaches. Models can also be used to
calculate flux values for regions where less intensive or no measurement data are available.
One of the challenges in model development is the integration of properties or variables in
space and time, accounting for the spatial and temporal variability of processes involved in
gas production, consumption and exchange.
Scaling not only comprises bottom-up approaches, but also top-down methods, such as
inverse modelling to calculate from the atmospheric concentrations back to the sources. Top-
down scaling in general involves the validation of estimates obtained at a lower scale level
against constraints given at a higher level of scale. Hence, scaling requires uncertainty analysis

between different scales ?; (ii) How can we best define functional types and integrate state
variables and properties in time and space ?; (iii) What is the relation between scale, the
model approach and the model parameters selected ?; (iv) How should the uncertainties in the
results of scaling be investigated ? The four group reports are included in this volume as
separate chapters together with the peer-reviewed background papers.
The organizing committee for the workshop, which started discussions Jn 1996, included
the following members: A.F. Bouwman (National Institute of Public Health and the
Environment, Bilthoven), N.H. Batjes (International Soil Reference and Information Centre,
Wageningen), H.A.C. Denier van der Gon (Soil Science and Geology Department, Wageningen
Agricultural University), F.J. Dentener (Institute for Marine and Atmospheric Research,
Utrecht University), J. Duyzer (TNO Institute of Environmental Sciences, Energy Research and
Process Innovation, Apeldoorn), W. Helder (Netherlands Institute for Sea Research, Den Burg),
J. Middelburg (Netherlands Institute of Ecology, Centre for Estuarine and Coastal Ecology,
Yerseke).
The organization of the workshop was made possible through funds of the Commission of
the European Communities (CEC-DG XII), European IGAC Office (EIPO), International
Fertilizer Industry Association (IFA), Kemira Agro Oy, National Institute of Public Health
and the Environment (RIVM), Norsk Hydro, Netherlands Royal Academy of Arts and
Sciences (KNAW), Shell Nederland b.v., and the Netherlands Organization for Applied
Scientific Research (TNO).
Cooperating organizations were the Intemational Society of Soil Science (ISSS),
International Geosphere-Biosphere Programme (IGBP), International Global Atmospheric
Chemistry Programme (IGAC), Global Emission Inventories Activity (GEIA), Centre for
Climate Research (CKO), and the Climate Change and Biosphere Programme of the
Wageningen Agricultural University (CCB)
Dr. L.R. Oldeman
Director, Intemational Soil Reference and Information Centre (ISRIC)
October 1998
ix
ACKNOWLEDGEMENTS

9 Elsevier Science B.V. All rights reserved
TOWARDS RELIABLE GLOBAL BOTTOM-UP ESTIMATES OF TEMPORAL
AND SPATIAL PATTERNS OF EMISSIONS OF TRACE GASES AND AEROSOLS
FROM
LAND-USE RELATED AND NATURAL SOURCES
A.F. Bouwman l, R.G. Derwent 2 and F.J. Dentener 3
~ National Institute of Public Health and the Environment, P.O. Box 1, 3720 BA Bilthoven, The Netherlands
2 Meteorological Office, London Road, Bracknell, RG 12 2SZ Berkshire, UK
3 University of Utrecht, Institute for Marine and Atmospheric Research, Princetonplein 5, 3584 CC Utrecht, The
Netherlands
I. Introduction
Emission inventories play a dual role in global air pollution issues. Firstly, they can be used
directly to establish the more important source categories, to identify trends in emissions and
to examine the impact of different policy approaches. Secondly, emission inventories are used
to drive atmospheric models applied to assess the environmental consequences of changing
trace gas emissions and concentrations and to provide advice to policy makers. This second
role contributes to the atmospheric modelling community being an important user of emission
inventories. The assessment process for global air pollution problems has a number of
identifiable steps: (i) it quantifies the changes in trace gas composition of the atmosphere; (ii)
it quantifies changes in atmospheric chemistry, transport, deposition and radiative forcing;
(iii) it identifies the climate responses to the changes in atmospheric composition of the
radiatively active trace gases; and, (iv) it quantifies the biological and ecological responses to
the predicted changes in climate.
The atmospheric modelling community will need a hierarchy of emission inventories to
complete an assessment of global air pollution problems based on these steps over the next
decade or so. In their simplest form, atmospheric models merely require no more than fixed
global emission fields of each relevant species. However, in their most complex form, future
atmospheric models will require emission fields whose spatial patterns and magnitudes will
respond in a wholly self-consistent manner to changes in economic prosperity, demography,
land use, climate change and policy. The requirements placed on the emission inventories will

MOGUNTIA Eulerian 10~ 10% 10 levels monthly CH4 and NOx 11
chemistry
MOZART Eulerian 2.8%2.8% 18 levels 6-hourly 45 species 12
RGLK Eulerian 10~ 10% 0 levels Monthly SO2, NOx and NH3 13
chemistry
STOCHEM Lagrangian 3.75~215 19 '. ~els 6-hourly 70 seecies 14
UiO/GISS Eulerian 8~ 10~ levels 8-hourly to 5-day 50 species 15
a 1, Li and Chang (1996); 2, Allen
et al.
(1996); 3, Moxim
et al.
(1996); 4, Chin
et al.
(1996); 5, Penner
et al.
(1994); 6,
Mt~ller and Brasseur (1995); 7, Roelofs and Lelieveld (1995); 8, Kraus
et al.
(1996); 9, Wauben
et al.
(1997); 10, Flatoy and
Hov (1996); 11, Dentener and Crutzen (1993); 12, Brasseur
et al.
(1997); 13, Rodhe
et al.
(1995); 14, Collins
et al.
(1997);
15, Berntsen
et al.

2. Some of the trace gases and aerosol species handled by current chemistry transport models (CTMs) for the
assessment of global air pollution problems.
Radiatively active gases Aerosols
CO2 Black carbon
CH4 Organic particles
N20 Wind-blown dust
CFCs: 11,12,113 Sea-salt
HCFCs: 22, 141b, 142b Resuspended material
HFCs: 134a, 152a Volcanic emissions
Perfluoro molecules: SF6, CF4, C2F6, C4F8 Biomass burning
Ozone precursor and depleting gases Aerosol precursor gases
CO SO2
NOx DMS
H2 H2S
Synthetic hydrocarbons: light C2 - C~o hydrocarbons, NH3
oxygenates
Biogenic hydrocarbons: isoprene, terpenes
CH3CC13, C2C14, C2HC13, CH2C12
CH3Br
CH3CI
Synthetic bromine compounds: 12B 1, 13B 1
1992) and global change research in general (O'Neill, 1988).
Two approaches are used for scaling gas fluxes: bottom-up and top-down scaling. Bottom-
up approaches, calculated from smaller to larger scales, involve extending calculations from
an easily measured and reasonably well understood unit to more encompassing processes. In
bottom-up scaling, various problems occur, such as how to aggregate the spatial and temporal
variation of properties or fluxes. Other problems are the various data uncertainties involved,
and translating mechanisms and processes between different scales.
Top-down approaches can mean using the measurements at a higher scale level which set
the boundary conditions for problem identification, and stimulate the testing of general

We will focus here on bottom-up scaling approaches for trace gas fluxes between aquatic
and terrestrial ecosystems (including agroecosystems) and the atmosphere used in the develo-
pment of global gridded emission inventories. The discussion will be primarily on emission
inventories prepared for scientific purposes such as atmospheric modelling. Although our
findings may also hold for other types of inventories, we will not discuss these inventories
explicitly on the country or provincial (sub-national) scale. Such inventories are now being
prepared for non-scientific purposes (e.g. national communications in the United Nations
Framework Convention on Climate Change).
The first, and major, part of this paper discusses the uncertainties and problems of
aggregation, generalization, stratification and modelling in the compilation of inventories.
Next, the available global emission inventories for land-use related and natural sources of
trace gases will be discussed on the basis of their spatial and temporal resolution. Finally, the
spatial and temporal resolution of current CTMs will be confronted with the available
emission inventories.
2. Uncertainties in emission estimates
Among the various approaches to estimating fluxes, the major ones in use are the emission
factor approach and modelling. In emission factor approaches emission estimates are derived
by combining measurement data with geographic and statistical information on the ecosystem
processes and economic activity. This can be represented as:
E= A • Eu
(1)
where E is the emission, A the activity level (e.g. area of a functional unit, animal population,
fertilizer use, burning of biomass) and
EU
the emission factor (e.g. the emission per unit of
area, animal, unit of fertilizer applied or biomass burnt). When using the emission factor
approach, both the stratification scheme for delineating functional types (e.g. management
systems, ecosystems, environmental provinces or entities) as a basis fol scaling, and the
reliability of the emission factor determine the accuracy of the flux estimates.
The firmest basis for scaling is developing an understanding of the mechanisms that

fluxes in field and laboratory experiments varied by more than a factor of 2 for most species
as a result of differences in fuel and burning conditions, one single emission factor was
proposed for each gas species, representing the aggregated flux for smouldering and flaming
fires for all fuel types (grass, wood, crop residues, etc.). For biomass burning it is difficult to
delineate the types of fires and the different techniques used may introduce systematic
differences, especially where reactive and difficult-to-measure species (such as NOx and NH3)
are involved. Clearly, one emission factor cannot describe all the burning conditions and fuel
types.
Another example illustrating the lack of measurement data concerns the emission
coefficients used for animal housings in Europe. In housings with mechan!cal ventilation the
gas flux can be easily determined from the gas concentration in the ventilation air and the flow
rate. The trace gas emissions from naturally ventilated housings can only be determined
indirectly and with greater uncertainty. In such "open" housings the emission depends on the
opening and closing of doors. In large parts of Europe, housings for cattle - the most important
category- are naturally ventilated (Asman, 1992). Besides being scant, the available
measurement data need not be representative. For example, the NH3 ammonia emission per
animal may vary by a factor of 4 within the same type of housing (Pedersen
et al.,
1996). This
may be caused by differences in the ventilation over the slurry between housings and by
differences in waste management practices such as cleaning.
Guenther
et al.
(1995) were also confronted by a lack of measurement data in their global
invemory of fluxes of volatile organic compounds (VOC) from vegetation. Measurements
represented only 26 of the 59 global land-cover types considered; the remaining land-cover
types, including tropical seasonal forests and savannas, were assigned an emission on the
basis of expert knowledge. In this database, most of the simulated VOC emissions come from
systems where very few or no measurements are available.
- Functional types. Guenther

have an enormous impact on the results of atmospheric models.
2.1.2. Regression approaches
Bouwman
et al.
(1993) calculated the N20 emission from soils under natural vegetation using
a simple global model describing the spatial and temporal variability of the major controlling
factors of N20 production in soils. The basis for the model is the strong relationship between
N20 fluxes and the amount of nitrogen (N) being cycled through the soil-plant-microbial
biomass system. The model calculates the monthly N20 production potential from five indices
representing major regulators of N20 production (soil fertility, organic matter input, soil
moisture status, temperature and soil oxygen status). These five indices were combined in the
final N20 index (Figure 1). Comparison of the N20 index with reported measurements for
about 30 locations in six ecosystems correlated with an r 2 of- 0.6 (Figure l a). The resulting
regression equation was used to calculate emissions on a l~ 1 ~ resolution. However, the
correlation coefficient is not a robust statistical method (see Sofiev, 1999), and minor
differences in only one of the measurement sites can cause major shifts in the correlation
coefficient (Bouwman
et al.,
1993). A major problem causing unreliability of the regression
equation is the lack of measurement data, particularly for a number of important ecosystems
and world regions that have not been sampled at all. It is not known how the model performs
in these areas (Figure 1 b).
2.1.3. Process models
Reliable regional or global estimates of trace gas emissions depend on an examination of
methodologies to reduce the current high uncertainty in the estimates. One potential way to
do this is to develop predictive flux models. Such models have been developed for different
processes and gas species on different scales. Examples will be given of the magnitude of the
uncertainties in global models, the value 9f models developed for speci~c ecosystems for
extrapolation and the problem of selecting the appropriate scale of process descriptions in
models. Finally, the advantages of using a range of models on different scales will be

o
i + ;
N20 index
b. Location of measurement
sites
90 90
-30 -
""" .,~. " - tc-' ~S'.: ~ .~
I"~ 1- : : ~, (~: .'%: " . "
~,.~-,,:,~: , :-~ ~ ~ ,: . ~ ,~) , ,,': . ~.~ ~,~ , -,,
" ;~ ' ~-~,~,,, - , ~=' "
.: ,~ , !
-:.:- ,~ .
9 . , :'~.',]:- , ' " ~: (., : _. -, ,,
:':
"~ : "-~ ~ " ~ "
-30
:: 6 .,.:
-60 - O
Measurement sites -
-60
, ,::':G- -
9 -
-90
-90
- -1 ,o 6 6'o
Figure 1. a) Relationship between measured N20 flux and simulated N20 index; and b) the location of the
measurement sites. Figure 1 a was modified from Bouwman et al. (1993) with kind permission from the American
Geophysical Union.
gas transfer velocity from global

soils on the field scale. The model can also describe NO• fluxes by using soil, climate and
data on management to drive three submodels (i.e. thermal-hydraulic, denitrification and
decomposition submodels). The management practices considered include tillage timing and
intensity, fertilizer and manure application, irrigation (amount and timing), and crop type and
rotation.
One of the processes simulated by the model is microbial growth. Since model results
appear to be dominated by the effect of temperature and soil moisture, which operate at nearly
all levels in the model, the question arising is whether there is an imbalance in the scales
according to which processes are described. The similarity of the results obtained for
shortgrass ecosystems by Mosier and Parton (1985) with their simple approach to those of Li
Simulated (g m "2)
75-
(a)
60-
45-
30-
15-
0 /
45
model correspondence
(rA2=0.905, n=36)
~
~o (:)
o
o o
O O
1"1 correspondence
(b)
model correspondence u~
(rA2=0.893, n=36)

production, oxidation and emission in flooded rice fields enabled them to develop a semi-
empirical model. They also derived a simplified (summary) version of the model for
application to a wider range of conditions but with limited data sets. Huang
et al.
(1998)
hypothesized methanogenic substrates as being primarily derived from rice plants and added
organic matter. Rates of methane production in flooded rice soils are determined by the
availability of methanogenic substrates a,~d the influence of environmemal factors. Model
validation against observations from single-rice growing seasons in Texas (USA)
demonstrated that the seasonal variation of methane emission is regulated by rice growth and
development. A further validation of the model against measurements from irrigated rice
paddy soils in various regions of the world, including Italy, China, Indonesia, Philippines and
the United States, suggested that methane emission could be predicted from rice net
productivity, cultivar character, soil texture and temperature, and organic matter amendments.
The detailed model and the summary model gave similar results (Figure 2), illustrating the
advantage of using simplified models.
2.1.4. Farm nutrient balance models
On the farm scale, trace gas fluxes occur in the stable, during grazing or during and after
spreading of animal manure. A model is therefore required to describe farm-scale processes
and cycles. For example, the model of Hutchings
et al.
(1996) describes NH3 losses from
animal housings, stored slurry, application of slurry and urine patches. The model builds on
knowledge acquired from various experiments and model studies of animal housing, waste
storage and farming practices. The model tracks the N input as animal feed until it is lost as
NH3. The problem of applying farm-scale models is the variety in management styles
occurring within groups of farms. Representative farms or averages for a group of farms have
to be used to obtain aggregated data. Differences in fluxes as a result of differences in
management may disappear due to this aggregation.
2.2. Uncertainties in the spatial distributions

corrected; this was due to the more complex relationships between precipitation and altitude.
- Uncertainty. The major problem is the inappropriate data coverage for large areas of the
world. The uncertainty of temperatures is particularly high in mountainous areas because there
are only a few weather stations in these regions and none of them are located on a clear
altitudinal gradient. The average moist adiabatical lapse rate for mountainous areas may result
in underestimation of temperatures for these areas. The spatial precipitation patterns resulting
from straight interpolation of measured values causes great uncertainty in areas with sparse
data coverage. Although the major annual cloud dynamics are represented, the regional
reliability of the cloudiness data is low.
2.2.2. Oceans
The best known chemo-physical global ocean data sets are included in the World Ocean Atlas
(Conkright
et al.,
1994; Levitus and Boyer, 1994a, b; Levitus
et al.,
1994). This database
includes spatial information on a l~ 1 ~ grid at various depths between 0 and 5500 m below
the surface for ocean temperature, salinity, dissolved oxygen, apparent oxygen utilization,
oxygen saturation, phosphate, and nitrate and silicate. Data for temperature and salinity have a
monthly time resolution and apply to depths between 0 and 1000 m below the surface; those
for dissolved oxygen, apparent oxygen utilization and oxygen saturation are on a seasonal
temporal scale and phosphate; nitrate and silicate concentrations taken on an annual basis.
-
Data limitations. The World Ocean Atlas is based on many observations. For example,
the temperature data set is based on 4.5 million profiles. Although the number of observations
is much higher than that used to produce the soil, vegetation/land cover and climate databases,
there is a problem of areas with a low density or absence of observations; furthermore, the
timing of the measurements may differ between profiles.
- Aggregation. The data at the observed depth were interpolated to standard depths. The
accuracy of the observed and standard level data was checked and flagged using a number of

not realistically describe the variability actually occurring within a soil unit in regions where
the density of observations is low.
2.2.4. Vegetation~Land cover
Similar to the soil information, land-use and land-cover information is required to scale up
information from the field to landscapes or ecosystems. Two examples of widely used
vegetation/land-cover maps are those compiled by Matthews (1983) and Olson
et al.
(1985)
with 1 ~ and 0.5 ~ spatial resolution, respectively. A recent development is the creation of a
global 1-km resolution global land-cover characterization (Loveland
et al.,
1997) based on
remotely sensed data. For the pan-European region (from Gibraltar to the Ural and from the
North Cape to Athens) a land-cover database with a 10% 10 minutes resolution was developed
(Veldkamp
et al.,
1996).
-
Data limitations. Matthews (1983) used the Unesco (1973) vegetation classification
scheme, while the database by Olson
et al.
(1985) is based on a land systems grouping.
Estimates of the extent of vegetation/land-cover types excluding cultivated land show a
considerable difference between the two databases. The global area of cultivated land is
similar in all the maps and corresponds well with FAO statistics, although regional
discrepancies may exist. The Olson and Loveland
et al.
databases include estimates for carbon
stocks in each land-cover type. Apart from definitional problems, there is generally a major lack
of observational data describing the properties of the vegetation/land-cover types distinguished.

and climate data may form an improvement here. The database also lacks data on the
characteristics of the vegetation type itself in the form of attribute data. Since the Loveland
et
al.
database is still in development, its uncertainty is as yet unknown. A review of the use of
remote sensing and other data in vegetation mapping is given by Estes and Loveland (1999)
2.2.5. Surrogate distributions
When the exact location or distribution of an activity or process is not known, surrogate
distributions are used to distribute activities, volumes or emissions over the grids. For
example, the grassland distribution is generally used to distribute cattle populations, while for
other animal categories the rural human population distribution or the distribution of arable
land is used as a surrogate distribution. However, the human population distribution is
generally not well known in rural areas, as statistics and atlases give data on populations in
major towns only. Using surrogate distributions may be realistic in some regions. However, in
others with specific stratifications of management, environmental or demographic conditions,
surrogate distributions may cause major errors (see, for example, the dairy cattle discussed in
2.4).
2.2. 6. General remarks
The major uncertainties in databases are generally related to the scarcity of data, and variable
density of data coverage and quality. With reference to the data problem, the mask files
(containing the number of grid points for data within the radius of influence surrounding each
grid box) provided in the ocean database form a good tool for describing the data density and
the point-by-point accuracy or reliability in other databases as well.
Compared to the classification schemes for vegetation and land cover in the traditional
maps and databases, satellite observations may provide a more flexible way of describing
ecosystem characteristics. Attribute files with descriptive data of the map units distinguished
(e.g. in the soil database of Batjes and Bridges, 1994) are very useful for modellers. These
attribute data also enable performance of statistical analysis of the data by unit. Furthermore,
correction of the satellite data with actual statistical information is a good way to improve the
accuracy of the spatial data. Finally, a combination of vegetation/land-cover data with climate

Data on crop production systems that are essential for estimating trace gas fluxes envelop
fertilizer use (including animal manure) by type and by crop, timing and mode of fertilizer
application, amount and timing of field-residue burning, animal waste management, number
of rice crops per year combined with soil and water management practices and fertilizer
application rates. Such data may be available for regions within countries but may not always
correspond to the official statistics or may be outdated.
Global forestry data are available from FAO statistics and assessments ~z.g. FAO, 1995).
However, information on the species planted and forest management are difficult to obtain. In
assessments of trace gas fluxes it is generally important to know the amount of above- and
below ground carbon in a certain forested area. Global data on carbon in vegetation can be
obtained from Olson
et al.
(1985), for example, and carbon in soils from such sources as
Batjes (1996).
In summary, the economic and attribute data generally have to be inferred from aggregated
country totals for the three land-use systems. Where the geographic distributions within
countries are not directly available, data have to be distributed over a spatial grid or
subnational regions. In this case surrogate distributions will have to be used (see section 2.2).
2.4. Uncertainties in the temporal distribution
Temporal patterns of trace gas fluxes vary in space. This poses difficulties for integration of
16
A.F. Bouwman, R.G. Derwent and F.J. Dentener
fluxes over spatial units. Spatial aggregation causes considerable loss of information on
temporal flux patterns. However, the paucity of measurement data often makes
generalizations unavoidable. Generalization is usually done by treating a landscape as a
composite of representative soils or farms with average waste characteristics, management
and weather conditions, or by treating populations as a group of identical members. Such
generalizations may lead to errors in temporal distributions due to averaging procedures. The
temporal pattern of estimates derived for a group of average farms may differ from the sum of
all individual farms. Generally, different grazing systems co-exist within regions. For

presented cropping calendars for rice production worldwide. This stratification serves as a basis
for applying flux models with the corresponding data on soil, water and crop management.
In summary, there is a problem in scaling-up of loss of information on temporal variability
due to spatial aggregation or generalization. This problem may occur on any scale.
Sophisticated and carefully chosen stratification schemes for the delineation of functional
types within landscapes may help in reducing the aggregation loss of information on temporal
variability. Temporal patterns can best be described by using process models.
3. Spatial and temporal resolution of current emission inventories and CTMs
3.1. Emission inventories
In the previous sections we discussed a number of major problems that occur during the pro-
cess of scaling-up data using different approaches on different scales. In this section we will
present a number of global and regional inventories for selected trace gas species and sources
of emissions which have been developed for scientific purposes. We will not discuss these
Towards reliable global estimates of emissions of trace gases and aerosols 17
Table
3. Global inventories of emissions of trace gases and aerosols from aquatic and terrestrial ecosystems for a number
of gas species with a spatial resolution of 1 o • 1 o longitude-latitude representative for the period around 1990.
Category
CO 2 CH4
CO VOC
N20 NO• NH3
S/SO• Aerosols Black
carbon
Land-use related sources
Crops, fertilized fields
Animals (including enteric
fermentation, animal
waste)
Biomass burning (including
waste and fuelwood com-

References: 1, Matthews
et al.
(1991); 2, Bouwman and Taylor (1996); 3, Yienger and Levy (I995); 4, Bouwman
et al.
(1997); 5, Lerner
et al.
(1988); 6, Fung
et al.
(1991); 7, Olivier
et al.
(1996); 8, Spiro
et al.
(1992); 9, Matthews and Fung
(1987); 10, Guenther
et al.
(1995); 11, Nevison
et al.
(1995); 12, Lee
et al.
(1997); 13, Benkovitz and Mubaraki (1996); 14,
Tegen and Fung (1995).
a Inventory based on estimates of burnt dry matter burnt can also be used for other gases.
b Inventory could be based on Bouwman
et al.
(1997).
c Inventory is in fact based on emission factors for biomes coupled with a mechanistic model to produce temporal patterns
of fluxes.
d Soil dust emissions and transport are simulated on the basis of GCM-based wind fields.
inventories on the country or provincial (subnational) scale being prepared for non-scientific
purposes (e.g. national communications in the United Nations Framework Convention on Cli-

for all known sources
States of America sources
Europe SO2, NO• NH3, VOC. CO for all 2~215 ~ grids (Ion. • lat.) Ya 4
known sources
The temporal resolution is indicated by y (year), or h (hour).
References: 1, EPA (1993); 2, EEA (1997); 3. UN (1995); 4. Veldt
et al.
(1991).
a with time profiles for conversion to monthly or shorter time periods
a simple model based on temperature and precipitation data from one particular GCM. Some
regional inventories include rules for distributing emissions in time, for example, on a daily or
hourly basis (Table 4).
National inventories will be produced in the framework in the IPCC Methodology for
National Inventories. Most of these inventories will be compiled on the basis of default annual
emission rates, as measurement data are not available in most countries. This temporal
resolution of one year is similar to that of most of the global inventories.
3.2. Atmospheric models
It is difficult to be definite about the current state of the art in CTMs since they continue to be
developed as scientific understanding grows and as computers increase in soeed and capacity.
Meteorological data with a time resolution of 1-6 h are typical of data used, while the spatial
resolution in the models is typically a few degrees latitude and longitude. Models have typical
runs of a few seasonal cycles: this is considered a mere snapshot when used for climate
calculations. Model processes are usually handled with the same spatial and temporal
resolution as the meteorological processes.
It is important for two main reasons to accurately assess the trace gas fluxes between
terrestrial and aquatic ecosystems and the atmosphere in CTMs. Firstly, CTMs need to
describe these trace gas fluxes realistically so as to accurately assess the trace gas life cycle on
the global or regional scale. Secondly, the CTMs may need to give an accurate representation
of the trace gas flux for a particular ecosystem or region. In the first case, the spatial
distribution of the flux may not be so crucial but it is important to achieve the correct total

the problems of scaling trace gas fluxes can be side-stepped with a simple parameterization.
Clearly, there is a huge gap in scale between the available dry deposition studies on the leaf or
canopy scale and the coarse grid squares of the CTM.
Wet deposition is a sporadic process which is difficult to describe adequately in models.
The coarse spatial resolution of the models is certainly an issue but perhaps more important is
their neglect of the detailed microphysical and chemical processes thought to be occurring in
rain clouds. Simulated global- or regional-scale wet deposition fluxes are available with reaso-
SOa concentration (ppb)
25 25
20 -
15-
10-
5 -
STOCHEM - -+- - Bridge Race - -11( - Lullington Heath
-O- - Ladybower x - Bloomsbun/ -q3- - Birmingham
~ - Sunderland - -0 - Bamsley - -U- - Cardiff 9
~ - Newcastle ~ - Leeds ~ - Bristol t~-
t- Liverpool al$ Birmingham East <) Hull .3'
x- Leicester [3 Southampton ~, Bexley ]r
O Swansea 9 Middlesborough " /
~" I I
N - i
9 ,.',_ .o /t ,x,, x [] o r F'.:, o
9 ~. \ ~ < ]~ ~ . ~ - "/i ".;~:.~"
>*.'-\ ii. ~.0"" " " " 9 '. '" "i" ~ '~
~. .~ x. ~_ ~ .# i
9 ~ .$1~'.'r, .~'. .~: 9
~ ~ .__ ~li
.Ik.^ ~ i~ ~


et al.,
1997). Resuspension of sea-
salts and wind-blown dust is often driven by high winds, which can be adequately represented
in CTMs. However, the state of the terrestrial surfaces, whether wet or recently ploughed, may
have a pronounced influence on resuspension, and these local factors are not often well-
defined on the coarse scales used in the CTMs.
3.3. Comparison of CTMs with emission inventories
With the exception of the spatial resoh".:.on of the emission inventories which meet the
requirements of current CTMs, there are major inconsistencies to remain between the CTMs
and the emission inventories which drive them. The most striking discrepancy between CTMs
and inventories is in the temporal scale, which is generally one year for the inventories and 1-
6 hours for the CTMs. Most CTMs include routines based on hypotheses on temporal flux
distributions at the model scale, or assumptions on temporal patterns are provided with the
emission inventories (see Table 4). Another way is to incorporate the trace gas flux model in
the atmospheric model, as done for example in some CTMs for NOx from soils.
For reactive species with short atmospheric lifetimes such as NH3, NOx and VOC, the
temporal scale gap is a more serious problem than for long-lived species. An additional gap
between inventories and CTMs is the number of VOC species; here, some of the mechanisms
describing the chemistry in CTMs require a much larger number of species than included in
current inventories.
A general major problem is that it is not always possible to ensure that consistent land use
and meteorological data are used throughout the modelling system including the emission
inventorie~. Furthermore, there are scaling problems with all aspects of CTI~ input data, some
of which are caused by limited computer resources, others by the focus of the modelling
system and yet others by lack of current understanding.
Turning to validation of emission inventories, the emission fields for long-lived trace gases
can be tested using CTMs on the basis of concentrations, trends, and seasonality and spatial
gradients of concentrations, as the chemistry is less crucial for long-lived species with fewer
fluctuations over the year. For other species, deposition rates can be used to validate model
results. A discussion of validation tools is, however, outside the scope of this paper. We refer

the heterogeneity. This is done, for example, by presenting frequency distributions for regions
or functional types, or the standard deviation for grid boxes,. In many cases the point-by-point
uncertainty is not known. However, even the indication of the maximum and minimum values
could be more helpful than the mean alone for sensitivity and quantitative uncertainty
analysis.
- Flux models.
Flux models should be used where possible to replace traditional emission
factor approaches. Firstly, models, which are descriptions of current process knowledge, are
preferred above simple rules such as those applied in CTMs to produce temporal distributions.
Secondly, intemal consistency of CTMs is improved by incorporating the flux models.
In trace-gas flux models there is often an imbalance between the level of detail by which
different processes are described. The relationship between scale, the model approach and the
model parameters selected is very important in this respect. On a higher scale the data
availability, generally poses a problem when using detailed process models. In this case,
simplified or summary models are expected to interpret field experiments with limited
information. The aim of simplifications is to make the model applicable to a wider area with
limited data sets. Developing such ranges of models from the micro-scale to field scale and
summary models to be used for extrapolation to other sites with different conditions is
extremely useful. Summary models will also help to develop a better understanding of how to
select the key variables to be used for specific scales.
- Environmental data.
The spatial data on climate, oceans, soils, land cover and land use
which are commonly used as a basis for scaling of trace fluxes have four general
characteristics: (i) their uncertainty is regionally variable but generally unknown in the spatial
distributions; (ii) data classifications are always aggregations (iii) classifications used are
generally not easily translated into other classifications; (iv) classifications cannot be easily
translated into properties or regulating factors of trace gas fluxes. In view of these
22
A.F. Bouwman, R. G. Derwent and F.J. Dentener
characteristics the use of common databases should be promoted.

ventories representing the long-range average have less value than time series of flux
estimates.
This paper has reviewed the uncertainties in estimating emissions from land-use-related,
and natural terrestrial and aquatic sources. A comparison has also been drawn up between the
available inventories and the requirements of state-of-the-art CTMs. We have shown a
number of weaknesses and problems in current methods for estimating emissions. We have
also presented several possibilities for improving flux estimates, hoping that these
recommendations will stimulate further study and discussion on the reduction of uncertainties
in flux estimates.
References
Allen, D.J., P. Kasibhatla, A.M. Thompson, R.B. Rood, B.G. Doddridge, K.E. Pickering, R.D.
Hudson and S-J. Lin (1996) Transport-induced interannual variability of carbon monoxide
determined using a chemistry and transport model. Journal of Geophysical Research 101:28655-
28669.
Andreae, M.O. (1991) Biomass burning: its history, use and distribution and its impact on environmental
quality and global climate. In: Levine, J.S. (Ed.) Global Biomass Burning. MIT Press, Cambridge, pp.
3-28.
Towards reliable global estimates of emissions of trace gases and aerosols 23
Asman, W.A.H. (1992) Ammonia emission in Europe: updated emission and emission variations. Report
228471008, National Institute of Public Health and the Environment, Bilthoven, Netherlands.
Batjes, N.H. (1996) Total carbon and nitrogen in the soils of the world. European Journal of Soil
Science 47:151-163.
Batjes, N.H. (1997) A world data set of derived soil properties by FAO-Unesco soil unit for global
modelling. Soil Use and Management 13:9-16.
Batjes, N.H. and E.M. Bridges (1994) Potential emissions of radiatively active gases from soil to
atmosphere with special reference to methane: development of a global database (WISE). Journal
of Geop.hysical Research 99:16479-16489:
Benkovitz, C.M. and M.A. Mubaraki (1996) Sulfur emissions from volcanic activity in 1985 and
1990. In: A.S. Lefohn, J.D. Husar and R.B. Husar (Eds.) Assessing historical global sulfur
emission patterns for the period 1850 and 1990. US Department of Energy, Washington DC, pp.

Davidson, E.A. and W. Kingerlee (1997) A global inventory of nitric oxide emissions from soils.
Nutrient Cycling in Agroecosystems 48:37-50.
Denmead, O.T., R. Leuning, D.W.T. Griffith and C.P. Meyer (1999) Some recent developments in
trace gas flux measurement. In: A.F. Bouwman (Ed.) Approaches to scaling of trace gas fluxes in
ecosystems. Developments in Atmospheric Science 24. Elsevier, Amsterdam, pp. 67-84.
Dentener F.J. and P.J. Crutzen (1993) Reaction of N205 on tropospheric aerosols: Impact on the
global distributions of NOx, 03 and OH. Journal of Geophysical Research 98:7149-7163.
EEA (1997) Corinair 94 summary report. European Topic Centre on Air Emissions. Report to the
European Environment Agency, I20 pp.
EPA (1993) Regional interim emission inventories (1987-1991) Report 454/R-93-021, U.S.
Environmental Protection Agency, Research Triangle Park.