Loading in 2 Seconds...
Loading in 2 Seconds...
Report on the joint TFMM/TFEIP scientific workshop Understanding discrepancies between atmospheric model results and measurements given uncertainties in emission inventories, models and measurements John van Aardenne (DG JRC)* and Peter Builtjes (TNO)
Understanding discrepancies between
atmospheric model results and measurements
given uncertainties in emission inventories, models and measurements
John van Aardenne (DG JRC)* and Peter Builtjes (TNO)
* views expressed are those of the writer and may not be regarded as stating an official position of the European Commission."
Increase the interaction between inventory, modelling and observation community to understand the uncertainties in the different tools for studies on air pollution in the EMEP domain.
Understanding where uncertainties need to be decreased or could be decreased best to better understand the air pollution system from source to effect.
Presented purpose and methodology on calculation of uncertainty in emission inventories and showed that one of the main purposes of performing uncertainty in emission inventories is to focus on the key sectors that require improvement.
Presented on overview of report uncertainty ranges:
Graph by Winiwarter (2007),
Sources: Rypdal, 2002; Schöpp et al., 2005; Keizer et al., 2006; Kühlwein&Friedrich, 2000; Leitao et al., 2007
Uncertainties in atmospheric observations: Wenche Aas (NILU/EMEP CCC)
Showed that uncertainty in measurement data is closely linked to data quality objectives of EMEP measurement network. These values do not cover errors due to spatial representativeness. Measurements are within DQO if reference methods are used and the site is representative, but there are exceptions.
DQO values can used as representative of uncertainty in the EMEP measurements:
Wenche Aas, TFMM/TFEIP Dublin 22 Oct 2007
Uncertainties in atmospheric modelling: Richard Derwent (Rdscientific)
Presented an uncertainty analysis in a photochemical trajectory model on a ozone event in which both emission inventory and model uncertainty where taken into account.
Concluded that the uncertainties in emission inventories where not the main source of uncertainty in the presented study. These where in this particular case, process descriptions and model parameterizations.
However, gridding, biogenic source and non-inventories emission like forest fires, agricultural burning and industrial fires will remain to have a large influence on ozone and PM episodes.
Showed consistent reduction trend in ozone precursors from different studies.
Showed as illustrative result the uncertainty by sector in the UK inventory clearly showing that sectoral uncertainty is larger than national total uncertainty.
Concluded that current emission inventories are good enough for integrated assessment at the European level but some sectors require more in-depth analysis: waste burning, soil-fertilization.
With large emissions source being reduced in the future the now minor sources (with large uncertainty) will become more important which might lead to higher uncertainty in the future emission values.
Showed that past decreases in high percentile ozone values in NW and Central Europe are clearly related to emission reductions.
Showed that changes in background ozone are not yet fully explained, but that hemispheric transport is important.
Illustrated that besides ensamble modelling, Monte Carlo Analysis can be used for model uncertainty assessment. In a urban ozone modelling exercise (Paris) the uncertainty was shown to be a factor of 2 and the result indicated that uncertainty in spatial and temporal resolution of the inventory was important.
One the question “Can we identify the reductions in ozone precursors from the increase in background concentrations” there was a consensus that the decreasing trend in peak ozone value could be explained by the emission inventories.
However, application of inventories in the study of urban ozone require a good spatial and temporal resolution.
There seem to be a need to study the role of soil NOx emissions which might become more important in the future given decreasing anthropogenic emissions. Also forest fires where mentioned as important source of emissions to be considered.
On uncertainty analysis in models using monte carlo analysis the usefulness was recognized but due to the complexity of the model, different assumptions might lead to the same results. Needs to be explored further.
Inclusion of voc speciation in the process description of the model was suggested as a way forward in modelling
Presented updated PM emission inventory and compared this with national inventories. For several countries differences where found.
Based on Monte Carlo analysis many of the differences with the countries where within the 95% confidence interval of the updated PM inventory.
For some countries the deviation was larger. Residential biomass burning or agriculture might be underestimated.
It seems that that no significant emission source is missing in the national reports.
Figure extracted from presentation Pulles (TNO) at TFEIP/TFMM workshop, Dublin, 2007
Recommended to check whether emission, models, and monitoring station study the same PM. Models do not always consider aerodynamic diameters whereas (emission and imission) measurements by principle do. Models calculate atmospheric concentrations, whereas measurements may be biased by sampling and analytical artifacts.
Suggested to compare SO42-, NO3-, OC, EC, etc… rather than PM. Models and measurements of PM may agree as a result of compensating errors. There’s no way to understand why models and measurements PM values differ if single major PM components cannot be compared.
Suggestion to check that what is called EC at emission sources (and transported, deposited, washed out by models) is the same as what is called EC at monitoring stations. It’s very probably not true, but both EC emissions and atmospheric concentrations can be translated to a common “reference EC” metric.
Modelling of particulate matter, Svetlana Tsyro (MSC-W/met.no)
Presented evaluation of uncertainties in primary PM emissions within the EMEP model
Results indicate that emissions of PM2.5 and PM10 are underestimated with respectively 23% and 28% in 2005.
Results suggest that EC/PM emissions from wood burning are underestimated in Central and Southern Europe.
Results suggest that EC/PM emissions of wood burning are overestimated in Northern Europe.
There might be a possible underestimation of emissions from road traffic and other mobile sources in central and southern Europe.
The research question structured the discussion but a conclusion was not reached (was not a criteria)
There is a need to explore to what extend the PM in emission factor measurements are comparable with PM as observed in the atmosphere.
There is no standard method to measure the PM emissions from residential stoves, also the purpose of use these stoves (heating/cooking vs. pleasure) leads different emission factors, which is currently not captured in the inventories.
It was mentioned that the uncertainty of fugitive sources such as agricultural activities is very high and this might be a significant source of pm emissions.
For scientific studies EC/OC speciation information in national inventories would allow for a better consistency check with the model-measurement results
Suggestion: emissions that are controlled by meteorology should be part of the model description except agricultural fugitive emissions
Presented significant uncertainties in current officially reported HM inventories due to missing sources/ re-suspension
One of the limitations found in the assessment is that sector detail in inventories do not provide information on HM relevant fuel and technologies.
Reported emissions of Hg seem to be more robust than those of other metals.
Main problem for validation and verification is the completeness in reporting, lacking a consistent dataset without gaps (need for independent estimates and gap filling).
Suggested to improved spatial and temporal resolution of heavy metal emissions for example by information on stack heights and other parameters for LPS
Presented different sources of emissions within the UK inventory.
In general point source emissions estimates seem to be well characterized but it might be that fugitive emissions are not included in the current LPS reports.
To improve inventory more comprehensive data is needed on metal content of fuels and a review is needed on brake and tyre wear information in existing literature.
Suggested to have natural sources incorporated into emissions inventory
Exchange of data between scientific communities is needed to improve inventories
Presented on model uncertainty analysis, model intercomparison and informatino from emission reporting.
Model uncertainty is 30-40% for Cd, Pb and 20-50% for Hg.
Current model-to-measurement comparison demonstrates 20-30% underestimation for Pb and 30-50% underestimation for Cd. Modelling results for Hg well agree with observations.
At the moment reported emission inventories for heavy metals are incomplete and of limited value in terms of model applications.
Model intrinsic uncertainty without effect of emissions.
On the research question “May we trust the model finding that emission inventories might underestimate heavy metals by a factor of 2-3 and if so, what is causing the underestimation in the emission inventories?
..the conclusion is that model results are robust.
To conclude on the cause of the uncertainty in the emission inventories we information on completeness of the inventories.
It is suggested that the modelling community provides an estimate of the missing amount of HM in inventories. However, currently there are not sufficient measurements available to apply inverse modelling techniques
Since NFR reported data does not provide the detail needed to calculate heavy metal emissions (e.g. metal content in fuel, fugitive emissions) there is a need for the provision of appropriate activity data (scientific project?).
There is a consensus between the different groups that collaboration is the way forward. Many suggestion where made on the different cases (see above) how to improve the quality of our toolbox. How to organize this and what priorities might exist was not discussed.
1. There is a need to link between model and inventories on which parameters should be provided by which community (e.g. missing emission sources, meteorological data needed for emission calculation (PM/NH3), speciation of compounds (VOC, PM)).
2. Non-inventory emission sources are in several cases an important source of uncertainty (biomass burning, natural emissions, soil NOx emissions).
3. Given the increasing demand for local air pollution studies there is a strong requirement for good spatial and temporal resolution of inventory data.
4. To make significant improvements in the scientific quality of heavy metal emission inventories an improvement in detailed activity data and emission factors is essential.
For PM and HM inventories there is a need for different (more) detailed data in comparison to traditional air pollutants and greenhouse gases.
Although no conclusions have been drawn on the cases presented (this requires further work, was not the aim of workshop)
we as chairs feel that
given overall model uncertainty of 30-40%
given overall measurement uncertainty (incl. spatial representation) is in the order of 30-40%
any discrepancy between model and observation significantly larger than 50% is probably the consequence of either uncertainty in the emission inventories or non-inventory emissions.
Participants of the workshop appreciated the format of the workshop and the information provided by the different communities.
We would like to recommended that this informal interaction activity becomes a returning event (~2 yrs).
Chair’s report will be send to participants early next week for 1 round of comments
Thanks again to all presenters for their contribution!