Uncertainties in emission estimates

Im Dokument Emission Inventories and Projections (Seite 113-117)

Chapter 3 Emission Inventories and Projections

3.2. Development of new emission datasets to study hemispheric transport of air pollution

3.2.5. Uncertainties in emission estimates

An uncertainty estimate is one of the quality indicators of an emission inventory and can be used to prioritise efforts to improve the inventory. Verification has been defined as:

the collection of activities and procedures conducted during the planning and development, or after the completion of an inventory that can be used to establish its reliability for the intended application of the inventory [IPCC, 2006].

Verification methods include comparisons of different inventories, comparisons of results of alternative methods, and using emissions in models to compare with atmospheric measurements.

These methods are complementary. Statistical approaches to estimate uncertainties in emission inventory levels and trends have been developed at large scale by the IPCC [2006] and in more specific applications [e.g., Frey and Zheng, 2002]. Two approaches are typically used: simple error propagation and Monte Carlo simulations. The main challenges in estimating inventory uncertainties are, however, uncertainty in the input data and developing methods to quantify systematic errors. For most inventory applications the random component of an uncertainty estimate will be small compared to the systematic component. The IPCC [2006] lists the following sources of uncertainties to consider:

lack of completeness, inventory model (estimation equation), lack of data, lack of representativeness of data, statistical random sampling error, measurement error, misreporting or misclassification, and missing data. Systematic expert judgments can be used to complement other sources of information on uncertainties. The usual metric for expressing uncertainty estimates is two standard deviations as a percentage of the mean.

In recent years there have been a number of new analytical tools applied to the elucidation of emissions emanating from sources in the northern hemisphere. Techniques include improved direct (forward) modelling and inverse modelling, making use of improved ground-station monitoring

0

88

networks, and aircraft observations during large-scale field campaigns. Also, a new generation of satellites has provided trends based on column data that have been compared with emission trends.

More often than not, the observation-based methods have suggested that emission estimates obtained from inventories, particularly for developing and newly industrializing countries, are too low. These important new tools and methods are described in section 3.6.

Uncertainties in inventories will vary by region, source, pollutant, and inventory year.

Uncertainty estimates for all world regions are not available. Generally it is expected that regions with the longest experience in compiling inventories and with well-developed statistical systems (e.g., Western Europe, North America and Japan) compile inventories with lower uncertainties than other regions. Formal quantification of uncertainties in emission inventories of non-greenhouse gas species has been performed in a few cases, e.g., Rypdal [2002] and the examples shown below.

The primary reasons for differences in uncertainties among sources are (i) activity statistics are missing or weak; (ii) emission factors and technologies are known better for some sources than for others; and (iii) emissions depend on natural and variable factors such as temperature and

precipitation. Usually, emissions related to the household sector, agriculture, and waste disposal are more uncertain than for transportation and large stationary sources. Natural sources and semi-natural sources (e.g., forest fires) are more uncertain than anthropogenic sources. The factors leading to uncertainty are many and varied. To mention just a few examples, emissions may be strongly

dependent on process conditions (e.g., burning vs. smouldering), on the level of abatement (e.g., new catalyst vs. broken catalyst), and on such unpredictable things as equipment vapour leakage. Data quality is strongly influenced by the extent of data available. For example, the speciation of NMVOC may be based on very few individual measurements that are extrapolated with dubious reliability to seemingly related emission sources that have not been measured. Laboratory and field testing of sources is an expensive proposition, so often the speciation of NMVOC emissions in China or India is based on measurements taken in Europe or North America, whether or not the sources can be

considered equivalent. Note that some emissions are more easily constrained (e.g., SO2 due to fuel S content, CO2 due to carbon content, and even Hg due to fuel/ore content) than others (e.g., CO emissions that depend on excess oxygen availability during combustion or the temperature-dependent emissions of semi-volatile PM).

Uncertainties for individual pollutants differ also with the level of experience of compiling an inventory, and these uncertainties typically can be reduced over time. SO2 inventories have a long history in Europe and North America and are considered relatively reliable in those regions. For other world regions, inadequate information about the sulphur content of fuels and sulphur removal

efficiencies may add to the uncertainty. NOx inventories are generally regarded as less certain than SO2 inventories, while NMVOC and CO inventories carry high uncertainties. Due to the short experience in compiling PM, BC, and OC inventories and the lack of data on the distribution of technology types in key regions, these are even more uncertain. In addition, PM emissions are very dependent on the abatement equipment installed and whether it is working as designed; the difference between a collection efficiency of 99% and 99.9% is a factor of 10 in the emissions. BC and OC inventories have uncertainty ranges from -25 % to a factor of two (higher for open burning) [Bond et al., 2004]. Typical reported ranges of uncertainty estimates for Europe are: SO2: ±5%, NOx: ±14%, NMVOC: 10-39% and CO: ±32% [EMEP, 2006]. The TRACE-P inventory [Streets et al., 2003]

estimated uncertainties in Asian emissions that ranged from ±16% for SO2 and ±37% for NOx to more than a factor of four for BC and OC. Within Asia, there was wide variation among countries and regions, with emission uncertainties in Japan being similar to those in Europe, and emissions in South Asia having high uncertainty. This topic is addressed further in our special review of Asian emission inventories in section 3.5.

A lower bound estimate of the uncertainty in the emission data used in the HTAP multi-model experiments discussed in the HTAP Interim Assessment Report and in Chapter 4 of this report can be obtained by comparing the emissions used in the modelling to the estimates from different

inventories. Figure 3.5 shows the EDGAR-HTAP 2000 emissions compared to the total emissions and anthropogenic emissions used in the SR1 experiments for 2001. The comparisons are provided for the

89

global total and each of the HTAP source regions. For the SR1 inputs, the means of the total and anthropogenic emissions are shown along with the maximum and minimum values for the ensemble.

While the ―anthropogenic‖ mean is calculated using data from all model runs, the ―total‖ mean is calculated only from model runs that included both anthropogenic and natural emissions. Some of the differences are due to emissions associated with land-use changes and savannah burning, which are not included in the EDGAR-HTAP 2000 data.

FINDING: Present-day emissions of most species important for intercontinental transport are relatively well understood by sector and world region. The spatial distribution of present-day emissions (gridded fields) is also reasonably well known.

RECOMMENDATION: Present-day emissions of some species are still unreliable in some parts of the world (e.g., black carbon and NMVOC emissions from developing countries, ammonia emissions, natural emissions), and additional resources are needed to measure the emission factors of key sources and to conduct surveys of activity levels. In particular, natural sources such as soil emissions, windblown dust, volcanoes, and remote biomass burning rarely fall within the purview of national governments and may need greater attention from the TF HTAP.

FINDING: Long-term emission trend datasets (century-scale) are becoming available and present a new opportunity to characterize intercontinental pollution flows in the past and future. Uncertainties are higher the further away we get from present-day conditions. Gridded emission distributions for the past and future are rudimentary.

RECOMMENDATION: Harmonization of past, present, and future emissions over time and space needs additional work. New spatially distributed proxy datasets are needed to more accurately distribute past and future emissions over the Earth’s surface.

FINDING: The reliability of emission data varies considerably by species, sector, and world region. This adds uncertainty to our ability to reliably model intercontinental transport among different world regions; it means that some source/receptor relationships are inherently better known than others because of inadequacies in our ability to quantify some of the source terms.

RECOMMENDATION: Greater emphasis needs to be placed on quantification of emission uncertainties. In addition, bringing together scientific communities working on emission inventories, satellite retrievals, laboratory tests, ground-based monitoring, and aircraft field campaigns could be a valuable function of the TF HTAP that would constrain emission estimates and narrow uncertainties. Making use of the local knowledge that goes into national or regional emission inventories can be helpful in improving global inventories.

90

Figure 3. 5. Comparison of the 2000 EDGAR-HTAP inventory to the emission inputs used in the SR1 run of the HTAP multi-model experiments representing 2001.

The comparison is presented for the global total and each of the four HTAP source regions. For the SR1 inputs, the means of the total and anthropogenic emissions are shown along with the maximum and minimum values for the ensemble. The EDGAR-HTAP 2000 data does not include any land use change emissions.

2000 EDGAR-HTAP

91

Im Dokument Emission Inventories and Projections (Seite 113-117)