Modelling of heavy metals and persistent organic pollutants: New ...

EMEP/MSC-E Technical Report 1/2010 July 2010

Modelling of heavy metals and persistent organic pollutants: New developments Ilyin I., A. Gusev, O. Rozovskaya, V. Shatalov, V. Sokovykh, O. Travnikov

DRAFT

msc-e Meteorological Synthesizing Centre - East Krasina pereulok, 16/1 123056 Moscow Russia Tel.: +7 495 981 15 66 Fax: +7 495 981 15 67 E-mail: [email protected] Internet: www.msceast.org

2

CONTENTS INTRODUCTION

3

1. SENSITIVITY TO METEOROLOGICAL PARAMETERS ASSOCIATED WITH CLIMATE CHANGE

5

1.1. Statement of the problem

5

1.2. Description of the method

8

1.3. Investigation of transport distance for POPs

9

1.4. Investigation of lead deposition

20

2. HEAVY METALS

26

2.1. Analysis of heavy metal model performance for 2008

26

2.2. Preparation of meteorological data for the EMEP case study on heavy metals

45

2.3. Comparison of modelled deposition and heavy metal concentrations in mosses

50

3. PERSISTENT ORGANIC POLLUTANTS

57

3.1. Comparison with measurements

57

3.2. Integrated monitoring/modelling/emission approach

66

3.3. Sensitivity of POP model to application of particle size-segregated deposition parameterization

72

CONCLUSIONS

78

REFERENCES

81

3

4

INTRODUCTION In 2010 the activity of MSC-E in the field of heavy metals and persistent organic pollutants (POPs) was focused on operational and research tasks. Operational tasks included assessment of pollution levels (concentrations and deposition) over the EMEP domain for 2008 basing on the available information on emissions, monitoring and modelling results. Besides, the modelling results were verified via comparison with measurement data. The uncertainty analysis of the components of pollution level assessment (emission inventories, monitoring data, modelling results) was carried out. Finally, contributions of transboundary transport to pollution of heavy metals and some POPs were established. Special attention was pied to the analysis of sensitivity of pollution levels to meteorological parameters associated with climate change. It is known that meteorological parameters significantly affect the levels and spatial distribution of HM and POP pollution levels across the EMEP domain. When annual changes of the emission values in countries are relatively small, e.g., in case of heavy metals, longterm variability of pollution levels is mostly controlled by variability of meteorological parameters. An increase of precipitation amounts favours higher wet deposition, and vice versa. Higher temperatures lead to faster chemical reactions of mercury oxidation and its consequent removal from the atmosphere, or to stronger decomposition of various. Changes in wind pattern can affect sourcereceptor relationships. Therefore, understanding of fate of atmospheric pollution levels in response to long-term climatic changes of meteorological variables is becoming actual and challenging task. The work aimed at linking of changes in climatic variables and pollution levels was initiated in MSC-E. In particular, the development of statistical approach based on application of regression analysis was started. Mosses in their tissues can accumulate heavy metals deposited from the atmosphere. MSC-E started the use of supplementary measurement information, such as data on concentrations of heavy metals in mosses, for the analysis of pollution levels in the EMEP region in 2009 and continued in 2010. More detailed comparison of the deposition of heavy metals and their concentrations in mosses was undertaken. In particular, the comparison was carried out for wet and dry deposition separately. Besides, the influence of wind re-suspension, period of accumulation, and density of moss sampling sites on the comparison was investigated. Significant efforts were focused on the evaluation of the HM and POP modelling results. First of all, modelled concentrations in air, in precipitation and wet deposition fluxes were compared with EMEP measurement data both on annual basis and on the shorter time scales (days, weeks, months). In order to explain temporal variability of the modelling results and find the relationship between modelled levels and emission sources back trajectory approach was extensively used. In particular, newly developed integrated monitoring/modelling/emission approach, based on back trajectory analysis, allows to link modelled concentration at each grid cell with the emissions in other grid cells through so-called influence functions. Pilot testing of this approach was performed for B[a]P at several EMEP stations. Discrepancies between the modelled and observed HMs and POPs quantities were discussed. However, in most cases close cooperation between national experts and the EMEP Centres is needed to investigate reasons of the discrepancies. 5

The cooperation between MSC-E, national experts and CCC can be achieved in the framework of the EMEP Case Study project. The main goal of the Study is the improvement of heavy metal pollution assessment over the EMEP region. The main attention will be paid to the analysis of the discrepancies between modelled values and measurements, using available information on emissions, monitoring and modelling results. In the framework of the project it is planned to use more detailed input data (emission, land-cover, meteorology etc.) with finer (e.g., 10×10 or 5×5 km2) spatial resolution and measurement data collected at national monitoring network together with data from the EMEP stations. Six countries (the Czech Republic, Croatia, the Netherlands, Italy, Slovakia and Spain) expressed their wish to participate in this activity. MSC-E has already started preparation of necessary input data for modelling. In particular, preparation of meteorological information is described in this report. MSC-E continued the research activity on the particle size–segregated description of removal processes. The current version of the MSCE-POP regional model treats particulate phase of POPs as mono-disperse fraction with fixed diameter. The main purpose of the research is to evaluate the feasibility of operational modelling of POPs atmospheric transport taking into account particle size distribution and to estimate the sensitivity of calculation results to the diameter of carrier particles. Several parameterizations of wet and dry deposition and different particle size distributions of B[a]P were investigated. The results of this research are available in this report.

6

1. SENSITIVITY TO METEOROLOGICAL PARAMETERS ASSOCIATED WITH CLIMATE CHANGE This chapter is focused on the development and testing of method aimed at linking pollution levels with meteorological variables associated with climate change. The applications of this method to investigation of transport distance of POPs (B[a]P, β-endosulfan) and deposition of lead in different parts of Europe are presented.

1.1. Statement of the problem The section is devoted to the investigation of the influence of meteorological and environmental parameters on the atmospheric transport of various pollutants (POPs and heavy metals). For the investigation of this influence for a given pollutant it is necessary to select: ¾

a target parameter describing the environmental pollution by the substance under consideration;

¾

a set of factors which are able a priori affect the target parameter.

The method applied is based on the regression analysis. For the application of this method both selected parameter and factors should be expressed as numbers. The input data is the set of values of the parameter p calculated on the monthly basis for various meteorological situations together with the corresponding values of factors fj, j = 1, … m. It should be stressed that the choice of factors is crucial for the results of investigation. From one hand, it is desirable that the set of the chosen factors contains all factors essentially influencing the investigated parameter, for the analysis to be complete. From the other hand, the number of the factors chosen should not be too large in order to ensure statistical reliability of their determination on the basis of the data available. Of course, the choice of the set of factors depends on the target parameter selected for the investigation and on the pollutant in question. Here two examples of application of the above approach are considered. They differ by selections of the target parameter and the factors influencing this parameter. One of important parameters for POPs arising in the evaluation of POP-like substances is their transport distance. So, as a first example, we consider transport distance TD of several POPs calculated on a monthly basis as a target parameter. We recall that TD is calculated on the basis of model simulations of POP transport from a conventional point source of standard power (1 t/y). The value of TD is defined as an average distance at which the pollutant concentration drops 1000 times in comparison to that near the source. The factors included into the investigation are temperature T, precipitation amount P and types of underlying surface over which the transport of the pollutant occurs (water bodies W, forest F and grass G). All these factors are enumerated as monthly means over the area covered by wind speed trajectories within the given month. To compare the dependence of the results on the meteorology of different years, calculations were performed for each month of four years (1996, 2001, 2003 and 2004). So in this case the input data for the analysis is the set {(TDi, Ti, Pi, Wi, Fi and Gi), i = 1, ..., 48}, where i is the month number.

7

Another target parameter which is of importance both for POPs and heavy metals is deposition flux to European countries. The investigation of the influence of meteorological parameters on this parameter is carried out for heavy metals. More exactly, the influence of meteorological parameters on monthly Pb deposition D to an European country (target parameter) was investigated as a second example of the application of the above method. Here friction velocity U*, precipitation amount P, temperature T, average wind speed V over the considered country and average wind speed Vtot over the territory of all European countries are considered as factors. Precipitation controls wet scavenging of particulate heavy metals like lead. Friction velocity is one of the key parameters affecting dry deposition. Country-averaged wind speed influence the amount of pollution emitted by national sources which are transported outside the country. European-scale wind speed characterizes transboundary transport to a considered country from other regions of Europe. It seems that the temperature itself cannot affect strongly the values of Pb depositions, but this parameter can influence the selected target parameter via the dependence with other meteorological parameters. For the analysis calculations of Pb depositions to four countries (the Czech Republic, Finland, Italy and the United Kingdom) for each month in the period from 1990 to 2007 were performed. So the input data for the analysis in this case is the set {(Di, U*i Pi, Ti, Vi, and Vtoti), i = 1, ..., 216}, where i is the month number.

1.2. Description of the method The method of investigation based on the application of regression analysis has been introduced last year in [Gusev et al., 2009], where the investigation of the influence of environmental parameters on POP transport distance was examined on the basis of one-year calculations for one reference year (1996). For the reader’s convenience, a brief description of the method is presented here. Let {pi, f1i, …, fmi}, i = 1, … N be a set of values of target parameter p complemented by the corresponding values of factors f1, ..., fm for various months i. This set can include both the entire set of input data described above and the subset of this set. Thus, in the investigation of POPs we shall use the set of data including all month of a given year. Under such an approach the influence of the considered factors to the target parameter within one and the same year is investigated. The results of such investigations for various years can be compared for revealing year-to-year differences in the influence of environmental parameters on the transport distance of the considered pollutants. On the opposite in the investigation of heavy metals (Pb) a set of months with one and the same name for all considered years is considered (for exmple, the set of Januaries of all years from 1990 to 2007). This approach allows revealing year-to-year differences for one and the same season. Both these approaches are tested below. The analysis is realized as step-by-step procedure. At the first step regression relations between target parameter p and each particular factor fj is constructed. The main factor is then chosen as the factor f j 0 for which the coefficient of multiple determination reaches maximum among all factors. At the second step the regression relation between p and pairs of factors (f j 0 fk ), k ≠ j 0 is considered. The factor of the second priority is again the factor for which the coefficient of multiple determination is maximal. The factors of each next level of priority are chosen in similar way.

8

1.3. Investigation of transport distance for POPs Here the results of the investigation of the dependence of the transport distance (TD) for two POPs, namely benzo[a]pyrene (B[a]P) and β-endosulfan on environmental parameters are described. To reveal this dependence, calculations of pollutant transport from conventional point sources located in Finland (30.1ºE, 62.7ºN), Italy (10.5ºE, 45.3ºN) and the United Kingdom (1.2ºW, 53.7ºN) were used. For the comparability of the results the intensity of each point source was taken one and the same – 1 t/y. For the comparison of the results for various years model simulations were done for four years (1996, 2001, 2003 and 2004). On the basis of these simulations monthly values of transport distance were evaluated for each substance and each year of simulations. The comparison of seasonal variations of TD for B[a]P predicted by the model for the considered years is presented in Fig. 1.1.

800 600 400

Dec

Oct

2001 2004 Sep

Jul

Jun

Apr

May

Mar

0

Aug

1996 2003

Nov

200

Jan

c

1000

Feb

Dec

Oct

Nov

Sep

Jul

Aug

Jun

0

Transport distance for B[a]P, km

2004

200

Apr

Dec

Oct

Sep

2001

2003

400

May

Nov

b

Jul

Jun

Apr

May

Mar

0

1996

600

Mar


Aug

200

800

Jan

2001 2004

400

UK source

1000

Feb

1996 2003

600

Jan

a

Italian source

800

Feb


Finnish source 1000

Fig. 1.1. Monthly based values of the TD for B[a]P calculated on the basis of simulations of B[a]P transport from Finnish (a), Italian (b) and British (c) sources, km The following peculiarities of TD seasonal variations can be marked: 1. Absolute values of the TD differ between chosen sources. Namely, the highest ability to longrange transport is calculated for Finland, and the lowest – for Italy (see Fig. 1.2). 2. There is a difference between values of the TD calculated for one and the same source in different years. The differences between the values of the TD calculated on annual basis for all source locations and all years considered are also seen in Fig. 1.2. 3. Seasonal variations of B[a]P transport distance for all sources and years have one and the same shape with higher values in cold months and lower values in warmer months. However, seasonal variations for each particular pollutant and source locations have their own peculiarities. Seasonal variations of the TD calculated on the monthly basis for β-endosulfan are presented in Fig. 1.2. Similar to the case of B[a]P, absolute values of the TD calculated for different sources are different (though the difference is les than for B[a]P). However, in contrast to the B[a]P case, in this case maximum TD value has been obtained for the United Kingdom (see Fig. 1.3). Another essential difference between these two pollutants is that the shape of TD seasonal variations is different for different source locations. Namely, for Italy and the United Kingdom the shape of the TD seasonal variations is similar to that obtained for B[a]P (higher values in cold months and lower – in warm ones). But for Finland the situation is the opposite: higher values of the TD are characteristic for warm months.

9

Italian source

300

c

2001 2004 Oct

Jan

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

0 Jan

b

1996 2003

100

Dec

200

Nov

100 Feb

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan

a

200

400

0 Feb

0

300

500

Sep

100

400

600

Jul

200

500

700

Jun

300

600

Apr

400

800

May

500

700

Mar

600

800

900

2001 2004

Feb

700

UK source

1996 2003

TD for b-endosulphane, km

800

2001 2004



900 1996 2003

Aug

Finnish source 900

Fig. 1.2. Monthly based values of the TD for β-endosulfan calculated on the basis of simulations of transport from Finnish (a), Italian (b) and British (c) sources

To explain the obtained differences in the TD values, it is reasonable to investigate the environmental factors affecting the ability of POPs to long-range transport. Here the factors affecting seasonal variations of the TD for the two above POPs will be examined by the above described method.

the UK

600

Finland

Italy

500 400 300 200

As described above, the following factors, which a priori can affect the TD for POPs, are included into consideration:

100 0 1996

1. Temperature T.

2001

2003

2004

Fig. 1.3. Annual based values of the TD for β-endosulfan for different sources and years

2. Precipitation intensity P.

3. Fractions of the area covered by trajectories emanated from a given point source occupied by water bodies (W), forests (F) and grassland (G). Below we present the results of the analysis for each of the two considered pollutants.

BENZO[A]PYRENE Table 1.1 summarizes the results obtained on the basis of calculations made with Finnish source for all considered years. Seasonal variations for this year are represented in Fig. 1.1a. The interpretation of the data in the table will be exemplified by the year 1996. Each of the four subsequent rows corresponds to steps 1 – 4 of the analysis and consists of: ¾

The list of parameters selected at the given step.

¾

Regression coefficients αT, αP, αW, αF, αG.

¾

The intercept S of the regression relation.

¾

Coefficient of multiple determination for the constructed regression relation.

10

Thus, at the first stage the temperature is chosen as the most priority parameter for explaining the variability of the transport distance. At this stage the following regression relation between the transport distance TD and temperature T is constructed: TD = αT · T + S = – 17.9 ·T + 600

(1)

approximating TD with coefficient of multiple determination 0.64. Table 1.1. The results of regression analysis of B[a]P transport distance calculated on the basis of the transport from the Finnish source for particular years Parameters involved 1996 T T, F T, F, G T, F, G, W 2001 T T, F T, F, P T, F, P, G 2003 T T, F T, F, G T, F, G, P 2004 T T, F T, F, G T, F, G, P

Temp. T

Precip. P

-17.9 -19.9 -21.7 -21.9 -19.1 -19.4 -19.5 -19.6 -8.7 -8.9 -9.4 -8.4 -10.1 -8.8 -9.6 -9.7

Regression coefficients Water Forest W F

1800

7.0 11.8

-40.9

-13.6

-3780 -4190 -2420

-4900 -4840 -4611

-2270 -2800 -2860

-1941 -2369 -2418

Grass G

Intercept S

Coefficient of multiple 2 determination R

2293 5140

600 1890 1160 -910

0.646 0.778 0.807 0.816

-834

560 2110 2080 2320

0.766 0.920 0.921 0.922

-2550 -2602

691 1436 2607 2705

0.403 0.510 0.563 0.584

2910 3160

710 1330 380 330

0.382 0.449 0.532 0.537

For further refinement of the TD seasonal variations a parameter of the second priority – forest fraction F – is selected. The corresponding regression relation is: TD = αT · T + αF · F + S = – 19.9 ·T – 3780 ·F + 1890

(2)

Application of this relation allows enlarging the multiple regression coefficient up to 0.78. The following two rows describe two factors of next priority (grass and water fractions G and W, respectively). The coefficient of multiple determination for the final regression relation is 0.81. The comparison of the initial TD for B[a]P calculated by the model for 1996 on the basis of the Finnish source with that calculated by the regression relations constructed at steps 1 – 4 is displayed in Fig. 1.4.

11

1200 TD

1000

Regression

Transport distance, km

800 600 400 200

400 200 Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan

1200 TD

1000

Regression


800 600 400 200

Regression

800 600 400 200 Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Step 4

Mar

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

0 Feb

0

TD

1000

Jan


600

Step 2

1200

Step 3

800

Feb

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan

Step 1

Regression

0 Feb

0

TD

1000

Feb


1200

Fig. 1.4. The comparison of the initial TD for 1996 with that calculated by the regression relations (Finnish source)

The comparison of the plots corresponding to different steps of approximation shows that at the first stage (consideration of the main priority parameter only) takes into account just a general tendency of the TD variations. Inclusion of the parameters of second and third priority subsequently improves the approximation. The inclusion of the fourth parameter leads to almost negligible improvement. This can be seen also from the consideration of coefficients of multiple determination R2 obtained at each approximation step (last column in Table 1.1). The inclusion of the second parameter changes this parameter from 0.646 to 0.778 whereas the inclusion of the fourth parameter leads to the minimum change of R2 from 0.807 to 0.816. The same is true for all other years also. It is seen that the approximation of the TD calculated by the model by regression relations is the worst for the year 2004. The relation can be improved by detalization of the set of parameters included into consideration (fractions of water, coniferous forest, deciduous forest, scrabs, grass, arable lands, bare soil and urban lands instead of water, forests and grass). With the help of these parameters the coefficient of multiple determination at fourth step becomes 0.681. However, the results of the analysis carried out with more detailed set of factors are less stable. For example, the factor of second priority under this approach will be the fraction of deciduous forest for three years (1996, 2001 and 2003) and coniferous forest for 2004. By this reason we restrict ourselves by the consideration of the set of aggregated parameters described above. Consideration of the list of priority factors shows that for all years the factor of the first priority is temperature followed by forest and grass fractions. The regression coefficients for these factors are negative, which shows that the ability of B[a]P to long-range transport decreases with the increase of the temperature and fractions of the area covered by vegetation. To examine the stability of ranking of factors between different years the analysis was applied to all 48 months of all considered years. The results are summarized in Table 1.2.

12

Table 1.2. The results of regression analysis of B[a]P transport distance calculated on the basis of the transport from the Finnish source for all years together Parameters involved

Temp. T -13.4 -13.4 -13.0 -13.0

T T, F T, F, P T, F, P, G

Precip. P

Regression coefficients Water Forest W F -2290 -2388 -2385

-23.6 -23.4

Grass G

-91

Intercept S 638 1386 1457 1490

Coefficient of multiple determination R2 0.466 0.532 0.540 0.540

The comparison of the results obtained for particular years with the result using all set of data shows that the two main parameters are the same for al considered sets of data: the temperature and the fraction of the area covered by forests. The contribution of changes in precipitation amounts occurring within the selected period are minimal. The plot comparing the variations of the TD with the results of calculations by regression relation taking in the account four priority parameters within four considered years is presented in Fig. 1.5.


1200 1000 800 600 400 200

TD

Regression

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

0

1996

2001

2003

2004

Fig. 1.5. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (Finnish source) The differences between the values of the TD calculated by the model and those obtained by the regression relation can be explained by several reasons. First, the set of factors chosen for the analysis may be incomplete. Second, regression analysis does not take into account the spatial variations of the parameters within the area covered by trajectories emanated from the source. However, regression relations explain about 55% of total variations of the TD, which is reasonable for such kind of the analysis. Now we proceed with the description of the results of the analysis obtained on the basis of simulations of B[a]P transport from the Italian source. Parameter ranking together with coefficients of multiple determination are shown in Table 1.3. It is seen that the results obtained for Italian source show higher multiple determination coefficient than that obtained for Finnish source. Coefficients of multiple determination for regression relations using four parameters range from 0.74 to 0.93 for the considered four years. The main priority factor is found to be grass fraction. Temperature is the factor of second priority for three from four considered years. The factor of the third priority is mostly water fraction. Precipitation and forest fractions are less important.

13

Table 1.3. The results of regression analysis of B[a]P transport distance calculated on the basis of the transport from the Italian source for particular years 1996 Coefficient of multiple Parameters determinati involved 2 on R G 0.635 G, T 0.760 G, T, W 0.821 G, T, W, P 0.826

2001 Coefficient of multiple Parameters determinati involved 2 on R G 0.764 G, T 0.855 G, T, W 0.900 G, T, W, P 0.928

2003 Coefficient of multiple Parameters determinati involved 2 on R G 0.878 G, W 0.889 G, W, T 0.899 G, W, T, P 0.900

2004 Coefficient of multiple Parameters determinati involved 2 on R G 0.460 G, T 0.579 G, T, F 0.686 G, T, F, W 0.735

The analysis carried out for all years together leads to the following ranking of the considered factors. The factor of main priority is grass fraction, followed by temperature, precipitation amount and forest fraction. The plot comparing initial TD with that calculated by regression relation including for listed parameters is shown in Fig. 1.6.


600 500 400 300 200 100

TD

Regression


0

1996

2001

2003

2004

Fig. 1.6. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (Italian source), km The coefficient of multiple determination for this approximation is 0.72. Thus, we see that, similar to the case of Finnish source, for Italy two main parameters are vegetation cover and temperature. Due to the peculiarities in the geographic location of the source, the type of vegetation mostly influencing long-range transport is grass. The results of the analysis obtained on the basis of simulations of B[a]P transport from the source of the United Kingdom are summarized in Table 1.4. Similar to the two above considered cases, temperature and vegetation cover are the two parameters of highest priority for almost all years considered. However, in the case of the United Kingdom, precipitation amount can be also a priority parameter for the description of the variability of TD. For two years (1996 and 2004) it is a parameter of the second priority. The consideration of multiple determination coefficients shows that in this case the inclusion of all four selected parameters leads to the improvement of the agreement between the TD values calculated on the basis of regression relation and the values calculated by the model. The values of multiple regression coefficient for relations using four factors ranges from 0.68 to 0.83 between the considered years.

14

Table 1.4. The results of regression analysis of B[a]P transport distance calculated on the basis of the transport from the British source for particular years 1996 Coefficient of multiple Parameters determinati involved on R2 F 0.503 F, P 0.675 F, P, T 0.784 F, P, T, W 0.811

2001 Coefficient of multiple Parameters determinati involved on R2 T 0.572 T, F 0.700 T, F, P 0.773 T, F, P, W 0.828

2003 Coefficient of multiple Parameters determinati involved on R2 T 0.569 T, F 0.632 T, F, W 0.664 T, F, W, G 0.676

2004 Coefficient of multiple Parameters determinati involved on R2 T 0.696 T, P 0.753 T, P, F 0.760 T, P, F, W 0.824

The results of the analysis obtained for all four years together show the following ranking of the factors. The factor of the first priority is temperature, then comes precipitation amount, forest fraction and water fraction in the order of importance. The coefficient of multiple determination for the relation including four parameters is 0.64.

1000 900 800 700 600 500 400 300 200 100 0

TD

Regression



The plot comparing initial TD with that calculated by regression relation including four listed parameters is shown in Fig. 1.7.

1996

2001

2003

2004

.. 1.7. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (the British source)

β-ENDOSULFAN The results obtained on the basis of calculations made with Finnish source for all considered years are summarized in Table 1.5. Seasonal variations of the TD are represented in Fig. 1.2a. It is seen that for three of four considered years main factors affecting the TD are temperature and forest fraction. For one of the years water fraction is considered to be a parameter of main priority. However, there is an essential difference between the results obtained for B[a]P and β-endosulfan. For the latter pollutant regression coefficients for temperature are positive. This means that the transport distance enlarges with the increase of temperature. The explanation of this fact may be as follows. The removal rate of a substance from the atmosphere is a crucial parameter determining the ability of this substance to be transported over long distances. This parameter is determined mainly by degradation and deposition to the underlying surface. Both these processes are temperature-

15

dependent. When the temperature is growing, the fraction of the gaseous phase of the pollutant increases, and degradation becomes faster. However, the rates of wet and dry deposition of gaseous phase decrease with temperature. So, the removal from the atmosphere is determined by two concurrent processes, and the overall removal rate depends on the relation between the rates of these processes. From the obtained results it follows that in the range of temperatures characteristic for Finland (typically from − 15 to +15 ºC) main removal processes are wet and dry deposition of gaseous phase which leads to the inverse character of temperature dependence of the TD (higher values of the TD in warm months, see Fig. 1.2a). We remark that for Italy and the UNITED KINGDOM, where positive temperatures prevail, regression coefficients for the temperature are found to be negative similar to the case of B[a]P. That is why the character of seasonal variations of the TD for βendosulfan is similar to that obtained for B[a]P (higher TD values in cold months).

Table 1.5. The results of regression analysis of B[a]P transport distance calculated on the basis of the transport from the Finnish source for particular years Parameters involved

Temp. T

Precip. P


Grass G

Intercept S

Coefficient of multiple determination R2

2126

1090 983 400 -1020

0.520 0.661 0.796 0.838

321

83 100 70 -76

0.582 0.867 0.880 0.884

-507 -757 -709

365 531 701 680

0.651 0.759 0.838 0.874

-809 -1690 -1839

346 605 1036 1117

0.458 0.632 0.733 0.748

1996 F F, T F, T, W F, T, W, G 2001 W W, T W, T, P W, T, P, F 2003 T T, F T, F, W T, F, W, P 2004 T T, F T, F, G T, F, G, P

2.4 3.6 4.0

798 1881

2.7 2.6 2.4

1218 1124 1200 1365

2.5 2.5 2.7 2.9 2.9 3.2 3.5 3.6

6.6 10.3

-13.0

-7.0

-2147 -1849 -630 497

-444 -329

-634 -732

Similar to B[a]P case the analysis was performed also on the basis of the data for all years together. The results are presented in Table 1.6. According to these calculations, first priority parameter is temperature, followed by forest fraction, water fraction and grass fraction. This is in a good agreement with the results of the analysis performed for each year separately. Regression coefficients for temperature obtained by these calculations are again positive which confirms the inverse character of seasonal variations of the transport distance (with higher values in warmer months).

16

Table 1.6. The results of regression analysis of β-endosulfan transport distance calculated on the basis of the transport from the Finnish source for all years together Parameters involved

Temp. T 2.9 2.8 2.8 2.7

T T, F T, F, W T, F, W, G

Precip. P


Grass G

-826 -601 -96

217 780

Intercept S 356 625 504 -193

1073

Coefficient of multiple 2 determination R 0.385 0.540 0.556 0.581

The coefficient of multiple determination obtained for the regression using all four parameters is 0.58. The obtained regression coefficients show that the last parameter is of a minor importance. The plot comparing initial TD with that calculated by regression relation including for listed parameters is shown in Fig. 1.8.


600 500 400 300 200 100

TD

Regression


0

1996

2001

2003

2004

Fig.1.8. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (Finnish source) The results of the analysis obtained on the basis of simulations of β-endosulfan transport from the Italian source are given in Table 1.7. The Table shows the parameter ranking together with coefficients of multiple determination. Table 1.7. The results of regression analysis of β-endoulfan transport distance calculated on the basis of the transport from the Italian source for particular years 1996 Coefficient of multiple Parameters determinati involved 2 on R G 0.813 G, T 0.827 G, T, F 0.832 G, T, F, W 0.835

2001 Coefficient of multiple Parameters determinati involved 2 on R G 0.793 G, P 0.868 G, P, W 0.902 G, P, W, T 0.911

2003 Coefficient of multiple Parameters determinati involved 2 on R W 0.863 W, P 0.891 W, P, T 0.905 W, P, T, F 0.907

2004 Coefficient of multiple Parameters determinati involved 2 on R G 0.407 G, F 0.636 G, F, P 0.707 G, F, P, W 0.761

Again the results obtained for Italian source show higher multiple determination coefficient than that obtained for Finnish source. Coefficients of multiple determination for regression relations using four

17

parameters range from 0.76 to 0.91 for the considered four years. The main priority factor is found to be grass fraction for most of the years similar to the result obtained during the investigation of B[a]P. The factor of second priority is temperature (for one of four considered years), precipitation amount (for two of considered years) and forest fraction (for one of the considered years). The factor of the third priority can be water fraction, temperature or precipitation. The analysis carried out for all years together leads to the following ranking of the considered factors. The factor of main priority is grass fraction, followed by temperature, forest fraction and water fraction. The plot comparing initial TD with that calculated by regression relation including for listed parameters is shown in Fig. 1.9.


400 350 300 250 200 150 100 TD

50

Regression


0

1996

2001

2003

2004

Fig. 1.9. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (Italian source) The coefficient of multiple determination for this approximation is 0.71. Thus, we see that, similar to the case of Finnish source, for Italy two main parameters are vegetation cover and temperature. In some meteorological situation water fraction can strongly affect the values of the TD. The results of the analysis obtained on the basis of simulations of β-endosulfan transport from the United Kingdom are summarized in Table 1.8. Similar to the two above considered cases, temperature and vegetation cover are the two parameters of highest priority for almost all years considered. However, in the case of the United Kingdom, precipitation amount can be also a priority parameter for the description of the variability of TD. For two years (1996 and 2004) it is a parameter of the second priority. The consideration of multiple determination coefficients shows that in this case the inclusion of all four selected parameters leads to the improvement of the agreement between the TD values calculated on the basis of regression relation and the values calculated bi the model. The values of multiple regression coefficient for relations using four factors ranges from 0.68 to 0.83 between the considered years. The results of the analysis obtained for all four years together show the following ranking of the factors. The factor of the first priority is grass fraction, then comes temperature, forest fraction and water fraction according to the order of importance. The coefficient of multiple determination for four parameters is 0.71. The plot comparing initial TD with that calculated by regression relation including four listed parameters is shown in Fig. 1.10.

18

Table 1.8. The results of regression analysis of β-endosulfan transport distance calculated on the basis of the transport from the British source for particular years 2001 Coefficient of multiple Parameters determinati involved 2 on R T 0.572 T, F 0.695 T, F, P 0.768 T, F, P, W 0.832

2003 Coefficient of multiple Parameters determinati involved 2 on R T 0.569 T, F 0.659 T, F, W 0.715 T, F, W, G 0.780

1000 900 800 700 600 500 400 300 200 100 0

TD

2004 Coefficient of multiple Parameters determinati involved 2 on R T 0.696 T, P 0.752 T, P, G 0.753 T, P, G, F 0.758

Regression



1996 Coefficient of multiple Parameters determinati involved 2 on R F 0.498 F, P 0.671 F, P, T 0.786 F, P, T, W 0.811

1996

2001

2003

2004

Fig. 1.10. The comparison of the initial TD for years 1996, 2001, 2003 and 2004 with that calculated by the regression relations (the British source)

Concluding remarks For B[a]P most valuable parameters determining its long-range transport are temperature and vegetation cover. The type of vegetation cover can be different depending on the geographical location of the source considered. For particular locations, precipitation intensity may essentially influence the ability to the long-range transport. For β-endosulfan, most valuable parameters determining its long-range transport are again temperature and vegetation cover. However, in contrast to the B[a]P case, the dependence of the TD for β-endosulfan on temperature is different for different temperature ranges. Namely, for lower temperature range (taking place in Finland) transport distance increases with the increase of temperature, which leads to the inverse seasonal variations of the TD (higher values of the TD in warmer months). For particular locations, water fraction may essentially influence the ability to the long-range transport.

19

1.4. Investigation of Pb deposition Here the results of the investigation of the dependence of lead (Pb) to the territory of four European countries (the Czech Republic, Finland, Italy and the United Kingdom) on environmental parameters are described. To reveal this dependence, calculations of Pb transport for the period from 1990 to 2007 were used. For the comparability of the results the same emission data were used. As an input data for the analysis depositions to the four above countries during each month of the considered period were calculated. The calculations were performed using MSCE-HM regional model. As stated above, the following five factors were included into consideration: friction velocity U*, precipitation amount P, temperature T, average wind speed V over the considered country and average wind speed Vtot over the territory of all European countries. To examine the influence of the chosen factors on deposition within one and the same season of all considered years twelve sets of input data were investigated: Januaries of each year, Februaries of each year, etc. Within each set of data priority factors affecting depositions to the territory of the chosen four countries were investigated. It should be noted that this investigation is the first attempt of the investigation of the influence of meteorological parameters on depositions of heavy metals and should be continued in future.

Czech Republic The results of the analysis are illustrated by depositions to the Czech Republic in January each year. The regression coefficients and coefficients of multiple determination obtained at each of five steps of evaluation are shown in Table 1.9. Table 1.9. The results of regression analysis of Pb deposition to the Czech Republic in January of each year in the period 1990 - 2007

Parameter s involved

V V, P V, P, T V, P, T, Vtot V, P, T, Vtot, U*

Friction velocity

Precip.

U*

P 0.27 0.26 0.31 0.20

Regression coefficients Wind Wind speed speed over Temp. over the European CZ countries T V Vtot - 2.8 - 3.7 - 1.6 -2.7 - 2.0 - 2.2 - 2.2

- 2.9

- 8.4

Intercept

Coefficient of multiple determination R2

S 116 118 98 159

0.52 0.59 0.63 0.66

137

0.68

It is seen that in this case the parameter of main priority is the wind speed averaged over the territory of the Czech Republic for the considered month. The parameter of the second priority is the precipitation amount over the country. Further follow temperature, wind speed averaged over the territory of all European countries and friction velocity. We remark that the temperature does not influence directly the deposition flux (at least according to the parameterization accepted in the model).

20

120

Pb depositions, g/km /month

Step 1

120

Step 2

100

2

100

2


Similar to the analysis performed for POPs, only three parameters of highest priority affect essentially the value of depositions. The influence of the last two parameters is much smaller. The agreement between depositions calculated by the model and those evaluated by the regression relation at all steps of evaluation process is shown in Fig. 1.11.

80 60 40 20

Deposition

Regression

60 40 20

Deposition

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Pb depositions, g/km /month

Step 3

120

Step 4

100

2

100

2


120

Regression

0

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

80

80 60 40 20

Deposition

Regression

60 40 20

Deposition

Regression

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

120

Step 5

100

2


1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

80

80 60 40 20

Deposition

Regression

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

21

100 90 80 70 60 50 40 30 20 10 0

Deposition

Regression

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

It is seen that poor correlation between model calculated depositions and those evaluated by regression relations can be explained by the fact that regression relation cannot reproduce peak values of deposition in May 1996 and May 2001. In the rest part of the plot regression curve agrees reasonably with evaluated one. The reasons of this disagreement should be investigated in future.

2

The coefficient of multiple determination obtained range from 0.57 to 0.78 depending on the chosen month. The only exception is the regression for Mays where the coefficient of multiple determination equals 0.34 (see Table 1.10 below). The comparison of model calculated depositions with those evaluated by regression relation at the last step of evaluation is presented in Fig. 1.12.


Fig. 1.11. The comparison of the Pb depositions to the Czech Republic in Januaries calculated by the model with that evaluated by the regression relations

Fig. 1.12. The comparison of the Pb depositions to the Czech Republic in May calculated by the model with that evaluated by the regression relations at the last step of evaluation

The summary of ranking of factors obtained for all months is given in Table 1.10. In the columns of this table ranks of the considered factors are presented by figures from 1 to 5; the minimal figure corresponding to the factor of higher priority. The last column of the table contains the average rank of each factor. The smaller is this figure, the higher the priority of the factors is. The factors are ranked by decrease of their priority. The last row in the Table contains the coefficient of multiple determination obtained with the use of all five parameters for each month. Table 1.10. The ranking of the factors by their influence on depositions to the Czech Republic for different months. The last column shows the average rank of each factor P V T U* Vtot R2

Jan 2 1 3 5 4 0.68

Feb 3 5 1 4 2 0.71

Mar 2 1 4 3 5 0.70

Apr 1 2 5 4 3 0.71

May 1 2 4 3 5 0.34

Jun 1 2 3 4 5 0.78

Jul 1 5 2 4 3 0.63

Aug 2 1 4 3 5 0.78

Sep 1 2 3 4 5 0.67

Oct 1 2 5 4 3 0.71

Nov 2 1 5 3 4 0.66

Dec 3 1 2 4 5 0.57

Aver 1.7 2.1 3.4 3.8 4.1

So, on the average, the factor of highest priority is the precipitation amount (P) over the territory of the Czech Republic. The factor of the second priority occurs to be the wind speed averaged over the country and over the considered month (V). Further follow air temperature over the country (T), friction velocity (U*) and the value of wind speed averaged over the territory of all European countries.

Finland

30


June

2

25

2


Similar investigation for Finland shows that the coefficient of multiple determination (R2) for most of the months range from 0.53 to 0.79. The exceptions are February and March with coefficients of multiple determination 0.18 and 0.41, respectively (see Table 1.11). The agreement between model calculations and predictions of regression relations for two months with maximum and two months with minimum R2 is shown in Fig. 1.13.

20 15 10 5

Deposition

Regression

20 15 10 5 Deposition

Regression

0

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

30

March

25

2


February

Regression

20 15 10 5

Deposition

Regression

0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Deposition

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

2


20 18 16 14 12 10 8 6 4 2 0

November

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

25

Fig. 1.13. The examples of the comparison of the Pb depositions to Finland for various months calculated by the model with that evaluated by the regression relations at the last step of evaluation

22

The disagreement between model calculations and predictions of regression relations may be conditioned by the incompleteness set of factors chosen for evaluation of depositions to the country. The summary of ranking of factors obtained for all months for Finland is given in Table 1.11. Table 1.11. The ranking of the factors by their influence on depositions to Finland for different months. The last column shows the average rank of each factor P T Vtot U* V R2

Jan 5 4 3 2 1 0.63

Feb 5 4 1 2 3 0.18

Mar 5 2 1 4 3 0.41

Apr 1 4 3 2 5 0.53

May 1 2 5 3 4 0.67

Jun 1 4 2 5 3 0.79

Jul 1 2 5 4 3 0.73

Aug 1 2 4 3 5 0.67

Sep 3 1 2 5 4 0.64

Oct 2 1 4 3 5 0.62

Nov 5 4 1 2 3 0.75

Dec 1 5 4 3 2 075

Aver 2.6 2.9 2.9 3.2 3.4

Similar to the case of the Czech Republic, the factor of main priority if precipitation amount. The interesting peculiarity of the result of regression analysis of factor influence for Finland is that the factors V and Vtot have changed their order of priority compared with the previously considered country. The value of the wind averaged over all European countries Vtot is included to the set of three factors of highest priority, whereas the average of the wind speed over the country V seems to be the factor of the lowest priority for Finland. This may be explained by the fact that depositions to Finland are in high extent determined by the transboundary transport.

Italy

The comparison of model predictions with evaluation by regression relation for May is shown in Fig. 1.14. Similar to the results obtained for the Czech Republic, the disagreement is conditioned by underestimation of depositions by regression relation for two years (1995 and 2001). The rest part of the plot regression curve reproduces depositions to Italy well enough.

140 120 100

80 60 40 20

Deposition

Regression

0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

2


The correspondence between model predictions and evaluation by the regression relation for Italy is much better than for other considered countries (similar to the case of POPs considered above). All multiple determination coefficients exceed 0.8 except for that for May equal to 0.74 (see Table 160 1.12).

Fig. 1.14. The comparison of the Pb depositions to Italy in May calculated by the model with that evaluated by the regression relations at the last step of evaluation

The summary of ranking of factors obtained for all months for Italy is given in Table 1.12. Similar to the previously considered countries, the factor of main priority is precipitation amount. The factors of the second and the third priority are the wind values averaged over the country and friction velocity. The value of the wind averaged over all European countries Vtot together with air temperature Tair are for this country factors of lowest priority. The ranking of the considered factors is close enough to that obtained for the Czech Republic.

23

Table 1.12. The ranking of the factors by their influence on depositions to Italy for different months. The last column shows the average rank of each factor P V U* Vtot T R2

Jan 1 3 4 2 5 0.87

Feb 1 3 4 2 5 0.88

Mar 1 5 2 3 4 0.84

Apr 2 1 3 5 4 0.81

May 1 5 2 3 4 0.74

Jun 1 2 4 5 3 0.85

Jul 1 2 5 4 3 0.97

Aug 1 2 4 5 3 0.83

Sep 1 3 2 5 4 0.81

Oct 1 3 2 5 4 0.84

Nov 1 2 5 3 4 0.81

Dec 1 4 2 3 5 0.84

Aver 1.1 2.9 3.2 3.8 4.0

The United Kingdom

50 45 40 35 30 25 20 15 10 5 0

2


March

Regression

40

May

35 30 25 20 15 10 5

Deposition

Regression

0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Deposition

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

2


The correspondence between model predictions and evaluation by the regression relation for the United Kingdom is good enough. The multiple determination coefficients for this country range from 0.62 to 0.81 except for March and May (see Table 1.12). The plots of the comparison of model predictions and regression evaluation of country depositions are shown in Fig. 1.15.

Fig. 1.15. The comparison of the Pb depositions to the United Kingdom in March and May calculated by the model with that evaluated by the regression relations at the last step of evaluation Similar to the results obtained above, the main part of the regression curve reproduces depositions to the United Kingdom well enough. Low values of multiple determination coefficient are conditioned by disagreement between model calculated depositions and regression relation for several particular years (1996 and 2004 for March and 1998, 2003 and 2006 for May). The summary of ranking of factors obtained for all months for Italy is given in Table 1.13.

Table 1.13. The ranking of the factors by their influence on depositions to the United Kingdom for different months. The last column shows the average rank of each factor V P T U* Vtot R2

Jan 1 5 3 2 4 0.69

Feb 1 2 3 4 5 0.70

Mar 1 5 2 4 3 0.50

Apr 1 2 5 4 3 0.65

May 2 1 4 3 5 0.54

Jun 1 2 3 4 5 0.62

Jul 5 2 1 3 4 0.62

24

Aug 2 1 3 4 5 0.80

Sep 2 1 3 5 4 0.64

Oct 2 3 1 4 5 0.65

Nov 1 4 2 3 5 0.66

Dec 1 2 3 5 4 0.81

Aver 1.7 2.5 2.8 3.8 4.3

For the United Kingdom, the factor of main priority is the wind velocity V inside the country. Precipitation amount P stands at the second place. The value of the wind averaged over all European countries Vtot is for this country a factor of lowest priority. This can be explained by the following two reasons. First, due to the geographical location of the United Kingdom, the transport from other countries affects the values of the country depositions in comparably low extent. Second, comparably high values of the wind speed lead to enhanced transport of the pollutant outside the country.

Conclusions So, it can be concluded that for two countries (the Czech Republic and Italy) the ranking of the considered meteorological factors has much in common. For both countries the two factors of highest priority are precipitation amount P (the factor of the first priority) and wind speed inside the country V (the factor of second priority). The value of the wind speed averaged over the territory of all European countries Vtot is a factor of low priority for both these countries. For the United Kingdom the two factors of the highest priority are again precipitation amount P and wind speed inside the country V; however, these two factors changed their places. Here V is the factor of the first priority, and P is the factor of second priority. This can be conditioned by high values of the wind speed in the United Kingdom which can lead to the enhanced transport of Pb outside the country. On the opposite, for Finland wind speed over the territory of all EMEP countries occurs to be one of the three factors of the highest priority whereas the wind inside the country is the factor of lowest priority. This phenomenon can be explained by the fact that the deposition to Finland is determined in high extent by the transboundary transport. It should be mentioned that the set of factors considered in the present investigation can be incomplete. This leads to the discrepancies between country depositions calculated by the model and evaluation of this parameter by regression relations in some months for the Czech Republic, the United Kingdom and, particularly, Finland. Further work is required for the improvement of this analysis.

25

2.

HEAVY METALS

This chapter is focused on the activities of MSC-E in the field of heavy metals. Pollution levels are described based on measurements and modelling information. Evaluation of modelled concentrations and depositions via comparison with measurements is presented. Quality air pollution assessment in the EMEP region was characterized using information on the uncertainties on emissions and measurements and the results of the model verification. Preparation and evaluation of meteorological data for the Case Study of heavy metal pollution assessment is overviewed. Application of concentrations of heavy metals in mosses for the assessment of pollution levels is discussed.

2.1. Analysis of heavy metal model performance for 2008 This section is devoted to the analysis of pollution levels of lead, cadmium and mercury based on modelling and monitoring information for 2008. Modelling results were obtained by MSCE-HM model. Measurement data related were submitted to CCC database by European countries. Spatial distribution of heavy metal deposition and concentration, verification of the modelling results as well as description of the pollution level changes between 2007 and 2008 are considered. Information on uncertainties of emissions, monitoring data and modelling results was summarized.

LEAD Over most part of the EMEP region total deposition of lead varied from 0.3 to 2 kg/km2/y (Fig. 2.1). Relatively high deposition levels (1-2 kg/km2/y) were associated with location of main sources of anthropogenic emission or wind re-suspension, the lowest deposition occurred in areas remote from regions with significant emission (Fig. 2.2). In addition to magnitude and spatial distribution of atmospheric sources of lead, deposition levels in Europe and Central Asia depends on meteorological peculiarities, first of all precipitation and wind patterns, and on properties of the underlying surface.

Fig.2.1. Spatial distribution of lead emissions over the extended EMEP domain in 2008

Fig.2.2. Total annual deposition of lead in Europe and Central Asia in 2008

In comparison with 2007, deposition levels in countries of Europe in 2008 declined. The most essential changes took place in southern, central and western Europe, where deposition declined by 15-30%. Smaller (5-15%) decline occurred in the eastern and south-eastern part of Europe. These changes were caused by several reasons. First of all, anthropogenic emission values of lead used in modelling for 2008 were lower than those for 2007. Besides, re-suspension in 2008 was also lower over most

26

part of Europe compared to 2007. In addition to this, in some parts of Europe precipitation amounts declined (e.g., Germany, Serbia, Bulgaria). In some countries, such as the United Kingdom and Ireland the deposition of lead rose between 2007 and 2008. The main reason for this is the increase of precipitation over this region, and in case of Ireland – also some increase of the emissions. Increase of deposition in the southern, eastern and south-eastern parts of Russia and in Central part of Kazakhstan was caused by increase of wind re-suspension flux in these regions. Besides, growth of precipitation amount in the central part of Kazakhstan favoured the increase of lead deposition. Both modelling results and monitoring data were used for the assessment of pollution levels over the EMEP region. Over major part of Europe and Central Asia annual mean concentrations of lead lay within 1-10 ng/m3 limits (Fig. 2.3). Relatively high concentrations took place in regions with significant emission sources where concentrations exceeded 10 ng/m3, and in some regions – even 20 ng/m3. In the central, western and northern parts of Europe both modelled and measured concentrations demonstrated similar pattern. In particular, in Germany, Slovakia and the Czech Republic the modelled and observed concentrations varied from 3 to 10 ng/m3, in Benelux region they could exceed 10 ng/m3 at some stations.

Fig. 2.3. Calculated and measured surface concentrations of lead in air over Europe and Central Asia in 2008

Relatively high calculated concentrations were indicated for the Balkan region, the eastern part of Ukraine and the European part of Russia. However, because of scarce measurement network in these regions it was not easy to confirm this finding. The lowest concentrations were obtained for Scandinavia and the northern part of Russia. Information on measured lead concentrations in air in 2008 was reported from 51 stations. The performance of the model was evaluated via comparison of modelling results with the most reliable monitoring data. Some of the reported measurement data were not involved in the comparison. The data from Spanish stations ES7, ES10, ES11, ES12 and ES13 cover very short period of observations (few days) and thus were not used in the analysis of pollution levels.

16 14

Observed

12

Modelled

10 8 6 4 2 0 AT2 AT48 AT5 BE14 CY2 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 DK10 DK3 DK31 DK5 DK8 EE9 ES8 ES9 FI17 FI36 FI37 FR13 FR9 GB13 GB17 GB91 HU2 IS91 LT15 LV10 LV16 NL10 NL8 NL9 NO1 NO42 PL5 SE14 SI8 SK2 SK4 SK6 SK7

Pb concentrations in air, ng/m

3

Relatively good agreement between modelled and measured values on annual level was found for stations located in Germany, Denmark, the United Kingdom, the Netherlands, Belgium, and Slovakia. For about 2/3 of stations the agreement between modelled and observed concentration is within ±50% (Fig. 2.4). For stations in Latvia, Norway, Finland, and Czech station CZ3 modelled concentrations exceed the observed ones by more than 50%.

Fig. 2.4. Comparison of modelled and measured annual mean concentrations of lead in air in 2008

27

Observed concentrations of lead at Hungarian station HU2 are very low compared to other stations. Annual mean levels are comparable with that measured at remote stations (e.g., FI36, DK10). For HU2 the model overestimated the observed concentrations by an order of magnitude. At stations in neighbouring countries such as Austria, Slovakia, Slovenia, the Czech Republic the modelled concentrations are of the same order of magnitude as the observed ones. Probably, this high overestimation is caused by some measurement problems [Aas W., personal communication]. For each station the comparison of measured and modelled time series was carried out. Special attention was paid to the analysis of the periods with relatively good agreement between the model and observations, and for identifying periods with marked discrepancies. For example, Fig. 2.5 demonstrates the comparison for the Dutch station NL8. As seen, the model managed to reproduce temporal variability of measured concentrations of lead in air for most part of the year. Most of measured maximums and minimums were captured by the model. Back trajectories approach was applied to link the modelled levels with emission levels in regions over which air masses were passing.

11-Dec

NL8

21-Nov

12-Oct

22-Sep

02-Sep

13-Aug

24-Jul

30-Jun

10-Jun

21-May

01-May

11-Apr

22-Mar

02-Mar

11-Feb

22-Jan

Observed Modelled (all sources) Modelled (anthrop)

01-Nov

90 80 70 60 50 40 30 20 10 0 02-Jan


3

Period between 17th and 21st of February was reproduced by the model. Map of back trajectory density shows that significant transport of air masses occurred through the territory of Belgium, southern Germany and eastern France (Fig. 2.6a). Belgium is known for high density of emissions (see Fig. 2.1). Therefore, elevated concentrations in the considered period can be explained, at least partly, by transport through territory of Belgium. Similar explanation can be applied for some other periods, e. g., October, 10th (Fig. 2.6b). As seen from the Figure, anthropogenic component of the modelled concentration in these periods is highly significant.

Fig. 2.5. Modelled and observed air concentrations of lead at the Dutch station Bilthoven (NL8) in 2008

b

a

Fig. 2.6. Density of back trajectories crossing grid cells for the Dutch station NL8 for the period of February 17-21 (a) and back trajectories for October 9 - 10 (b)

28

Some peaks (e.g., middle of April, middle of December) of the modelled concentrations overestimated the observed levels by an order of magnitude. These peaks were caused by wind re-suspension, which was quite high in the Benelux region compared to other regions (Fig. 2.7). Most likely it was overestimated for these periods in the considered region. However, for the major part of the year the re-suspension favours better agreement between modelled and measured values of concentrations.

Emissions, g/km2/day

Trajectories

Fig.2.7. Back trajectories crossing grid cells for the Dutch station NL8 for the period of 19th of April and overall (anthropogenic emission and wind re-suspension) of atmospheric input of lead

6 Observed Modelled (all sources) Modelled (anthrop)

5 4

ES9

3 2 1 31-Dec

26-Nov

23-Oct

20-Sep

27-Aug

26-Jul

24-Jun

08-May

07-Apr

28-Feb

30-Jan

0 04-Jan


3

For some stations air concentrations of lead were overestimated by the model. The comparison performed for weekly mean modelled and measured values at station ES9 (Spain) demonstrated that in some periods, mainly in summer and autumn, the modelled and measured concentrations agreed, while in the beginning and the end of the year the model often overestimated the observed values (Fig. 2.8).

Fig. 2.8. Modelled and observed air concentrations of lead at Spanish station Campisabalos (ES9) in 2008 Separate periods of measurements were analysed using back trajectories. High modelled concentrations of lead took place in the period from 30th of January to 3rd of February. Modelled concentrations exceeded the observed ones by 7 times. Back trajectories in this period were coming mostly from the northern part of the country (Fig. 2.9a). It is important to mention that the contribution of wind re-suspension was relatively small in this period (Fig. 2.10a). Among the anthropogenic sources the major contributor is Spain (Fig. 2.10a), which contribution exceeded 70% in this period. Precipitation amounts in the considered period were low compared to other periods. Therefore, lead emitted by Spanish sources was not washed out and thus modelled concentrations were high compared to other periods. Similar situation took place in the period from 23rd to 30th of December (Fig. 2.9b, 2.10b).

29

In the summer and autumn periods the agreement between modelled and measured concentrations was satisfactory. For example, the peak in the period 20th – 27th of September was well captured by the model. Main contribution to concentrations in air was done by Spanish sources (Fig. 2.10c).

a

b

c Fig. 2.9. Density of back trajectories crossing grid cells for the Spanish station ES9 (left) and total input of lead to the atmosphere (anthropogenic emission and re-suspension) for the periods 30th of January – 3rd of February (a), 23rd – 30th of December (b), and 20th – 27th of September (c) 9

Spain

3

Other

5 4 3 2 1

1 0.5 Sep, 27

Sep, 26

Sep, 25

c

Sep, 24

Dec, 30

Dec, 29

Dec, 28

Dec, 27

Dec, 26

Dec, 25

Dec, 24

Feb, 3

Feb, 2

Feb, 1

Jan, 31

Jan, 30

b

2 1.5

0 Dec, 23

0

0

a

6

3

Other Non-EMEP

2.5

Sep, 23

1

Resusp

Sep, 22

2

7

Sep, 21

3

Non-EMEP

Sep, 20

4

Spain Resusp

3.5

8

Conc. of Pb in air, ng/m

5

3

Non-EMEP Resusp Other Spain



3

6

Fig. 2.10. Contributions of various sources to modelled concentrations of lead in air at station Campisabalos (ES9) in the periods 30th of January – 3rd of February (a), 23rd – 30th of December (b), and 20th – 27th of September (c)

30

90 80 70 60 50 40 30 20 10 0

Observed

15-Dec-2008

01-Dec-2008

10-Nov-2008

27-Oct-2008

13-Oct-2008

22-Sep-2008

08-Sep-2008

25-Aug-2008

07-Jul-2008

23-Jun-2008

09-Jun-2008

26-May-2008

12-May-2008

28-Apr-2008

14-Apr-2008

24-Mar-2008

25-Feb-2008

11-Feb-2008

Modelled

14-Jan-2008

Precip sum, mm / period

Since the station is located in the centre of Spain, most of trajectories were passing over the country (Fig. 2.9) and in most of the periods the contribution of Spanish sources to modelled concentrations was dominating. Observed precipitation amounts was reproduced relatively well (Fig. 2.11). However, in some periods of time the model well matched the observations, and in other - overestimate them. Even if wind re-suspension was switched off, the modelled annual mean concentrating still exceeded the observed one by about 50%. There are several possible reasons of the discrepancies, which require further investigation. First of all, the model might underestimate wet scavenging or dispersion in the considered region. Another reason is spatial distribution of the emissions over the county. It is possible that in some regions of the country the emissions could be overestimated. In the framework of the Case Study it is planned to perform source-receptor modelling using administrative regions or big cities in a country as separate emission sources.

Fig. 2.11. Modelled and observed precipitation amounts at Spanish station Campisabalos (ES9) in 2008 Another measurable parameter was wet deposition flux. The combined map of modelled and measured wet deposition fluxes of lead in 2008 is demonstrated in Fig. 2.12. Since periods of observations at stations often did not cover entire year, for comparability the modelled and observed wet deposition fluxes were depicted as daily mean sums. Spatial distribution of modelled wet deposition first of all reflected distribution of the emissions (Fig. 2.1) and precipitation patterns (Fig. 2.13). In particular, relatively high deposition occurred in the central part of Russia, the eastern part of Ukraine, southern Poland, Slovakia, northern Italy, the Benelux region, and over the Balkans. Comparatively low deposition took place in parts of the EMEP region remote from main emission sources: in the Arctic, over Scandinavia, Iceland. Besides, low deposition was found over Africa and the desert regions of Central Asia. Measured deposition mostly followed this pattern. Relatively high (1.5 – 5 g/km2/day) deposition was indicated for Slovakia and the Benelux region. In Germany, Denmark, Poland and France modelled and measured wet deposition mostly lay within 0.8-2.5 g/km2/day. Smaller wet deposition fluxes were measured in Finland and Iceland. However, in some regions the measured values considerably exceeded the modelled ones. For example, these regions were Spain, Latvia, and the southern part of Norway. Low (0.2 – 0.8 g/km2/day) wet deposition of lead over the central part of Kazakhstan were confirmed by the monitoring data were submitted to MSC-E by national experts from this country. (Fig. 2.12). In the south-eastern mountainous part of the country the deposition exceeded 2.5 g/km2/y because of higher precipitation (Fig. 2.13) and emission values (Fig. 2.1). Comparison of modelled and measured concentrations of lead in precipitation and wet deposition fluxes in 2008 is demonstrated in Fig. 2.14 and Fig. 2.15. Good agreement between modelled and measured parameters was noted for stations in Germany, the United Kingdom, Denmark, the Netherlands, Poland, Sweden, Slovenia, Iceland and France. Both modelled concentrations in precipitation and wet deposition at most of stations in these countries matched the observed values 31

with bias within ±50%. The observed levels were underestimated at Spanish, most of Finnish and Norwegian sites. For some stations the agreement between calculated and observed concentrations in precipitation was satisfactory, but it was not the case for wet deposition (e.g., NO1), and vice versa (e.g., LT15). The reason for this is large difference between the modelled and observed precipitation amounts. Similar to concentrations in air, at about 2/3 of stations the difference between the modelled and observed concentrations in precipitation or wet deposition is within 50%, and for 1/3 – within 30%.

Fig. 2.12. Calculated and measured daily sums of wet deposition fluxes of lead over Europe and Central Asia in 2008

Fig. 2.13. Annual precipitation amounts in 2008

Concentration in precip, μg/L

4.5 4 3.5

Observed Modelled

3 2.5 2 1.5 1 0.5 BE14 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 DK20 DK22 DK31 DK5 DK8 ES8 ES9 FI17 FI22 FI36 FI53 FI8 FI92 FI93 FR13 FR9 FR90 GB13 GB17 GB6 GB91 HU2 IS90 IS91 LT15 LV10 LV16 NL9 NL91 NO1 NO39 NO55 NO56 PL4 PL5 SE51 SE97 SI8 SK2 SK4 SK6 SK7

0

Fig. 2.14. Modelled and measured annual mean concentrations of lead in precipitation at individual stations in 2008

4 3.5

Observed Modelled

3 2.5 2 1.5 1 0.5 0 BE14 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 DK20 DK22 DK31 DK5 DK8 ES8 ES9 FI17 FI22 FI36 FI53 FI8 FI92 FI93 FR13 FR9 FR90 GB13 GB17 GB6 GB91 HU2 IS90 IS91 LT15 LV10 LV16 NL9 NL91 NO1 NO39 NO55 NO56 PL4 PL5 SE51 SE97 SI8 SK2 SK4 SK6 SK7

2

Wet deposition fluxes, kg/km /y

4.5

Fig. 2.15. Modelled and measured annual wet deposition flux of lead at individual stations in 2008

32

50 45 40 35 30 25 20 15 10 5 0

Observed Modelled Anthrop

CZ3

01-Jan 15-Jan 29-Jan 12-Feb 26-Feb 11-Mar 25-Mar 08-Apr 22-Apr 06-May 20-May 03-Jun 17-Jun 01-Jul 15-Jul 29-Jul 12-Aug 26-Aug 09-Sep 23-Sep 07-Oct 21-Oct 04-Nov 18-Nov 02-Dec 16-Dec 30-Dec

Pb deposition flux, g/km

2

Modelled wet deposition of lead at station CZ3 was underestimated compared to measured values. Comparison of temporal variability of measured and observed wet deposition indicated that for some periods the model well matched the observed values while for other periods the underestimation was essential (Fig. 2.16).

Fig. 2.16. Modelled and observed concentrations of lead in precipitation at Czech station Kosetice (CZ3) in 2008 Back trajectory analysis was applied to the periods when the modelled values agreed well with measurements and when the discrepancies between measured and modelled values were high. For example, in the period from 18th to 24th of November the model significantly (around 3.5 times) underestimated measured pollution fluxes at CZ3. In this period and also other periods in October and November (e.g., 28th of October – 3rd of November) the prevailing transport of air masses was from the west (Fig. 2.17a) or south-west (Fig. 2.17b). The most important anthropogenic contributors to the deposition in the October period were Italy and Austria (Fig. 2.18a), and in the November period Germany and the Benelux region (Fig. 2.18b). When the eastern transport from territory of Slovakia or Poland took place (e.g., 16th – 22nd of September), the levels agreed satisfactory (Fig. 2.17c). As seen from Fig. 2.18c, the contribution of Polish sources ranges from 45 to almost 80%. Besides, on 20th of September role of Russian sources is substantial (around 20%). Precipitation amounts derived by meteorological pre-processor and observed at the station were similar in the October and November periods (Fig. 2.19). Uncertainties of the emission data could contribute to the significant discrepancies noted for October and November. However, it seemed unlikely that total emissions of main contributors – Italy, Austria, Germany, etc or of national sources was underestimated by the order of magnitude. Besides, there were some episodes (e.g., 12th – 18th of August) when the model well capture measured values, and the transport of air masses took place from the south-west. Probably, these high peaks of measured values were explained by the influence of local sources, or by temporal variability of national or external emissions not included in modelling. Case Study on heavy metal pollution assessment is aimed at integrated analysis of factors affecting agreement between modelled and measured values. Detailed analysis of emission, measurement data and modelling results will be carried out in cooperation with experts from the Czech Republic.

33

a

b

c Fig. 2.17. Density of back trajectories crossing grid cells for the Czech station CZ3 (left) and total input of lead to the atmosphere (anthropogenic emission and re-suspension) (right) for the periods 16th – 22nd of September (a), 28th of October – 3rd of November (b), and 18th – 24th of November (c)

5

Other Italy Non-EMEP

6

1

c

Nov, 24

Nov, 3

Nov, 2

Nov, 1

Oct, 31

Oct, 30

Oct, 29

b

2

Other Benelux

0

0 Oct, 28

Sep, 22

Sep, 21

Sep, 20

Sep, 19

Sep, 18

Sep, 17

Sep, 16

1

Czech Rep. Germany Resusp

Nov, 23

1

2

3

Nov, 22

2

3

Nov, 21

3

4

Nov, 20

4

4

Nov, 19

5

5

Nov, 18

6

0

a

Czech Rep. Resusp Austria

6

Wet dep. of Pb, g/km2/day

Other Russia

2

2

Wet dep. of Pb, g/km /day

7

Wet dep. of Pb, g/km /day

Czech Rep. Resusp Poland

8

Fig. 2.18. Contributions of various sources to modelled concentrations of lead in air at station Kosetice (CZ3) in the periods 16th – 22nd of September (a), 28th of October – 3rd of November (b), and 18th – 24th of November(c)

34

Observed Modelled

50 40 30 20 10

30-Dec-2008

16-Dec-2008

02-Dec-2008

18-Nov-2008

21-Oct-2008

04-Nov-2008

07-Oct-2008

23-Sep-2008

26-Aug-2008

09-Sep-2008

29-Jul-2008

12-Aug-2008

15-Jul-2008

01-Jul-2008

17-Jun-2008

03-Jun-2008

20-May-2008

22-Apr-2008

06-May-2008

08-Apr-2008

25-Mar-2008

11-Mar-2008

26-Feb-2008

12-Feb-2008

29-Jan-2008

15-Jan-2008

0

01-Jan-2008

Precip sums, mm / period

60

Fig. 2.19. Modelled and observed precipitation amounts at Czech station Kosetice (CZ3) in 2008

CADMIUM Total deposition of cadmium in 2008 ranged from 5 to 50 g/km2/y over most part of Europe and Central Asia (Fig. 2.20). In most polluted areas (south of Poland, the FYR of Macedonia, central regions of Russia) the deposition exceeded 100 g/km2/y, which is connected with the spatial distribution of emissions across the EMEP domain (Fig. 2.21). The lowest levels were typical for the Scandinavian countries because their national emission sources were relatively low and these countries were remote from main emission sources. Low deposition in the Central Asian counties such as Uzbekistan and Turkmenistan are explained by low precipitation amounts typical for arid regions. In comparison with levels assessed for 2007, deposition in 2007 declined in most of countries. Similar to lead, the decline was caused by various reasons, such as smaller wind re-suspension in 2008, and by lower values of cadmium emissions in countries. Significant reduction (20-30%) occurred in the western, southern and central part of Europe (France, Germany, Spain, Italy etc). Small growth of deposition in Ireland and on the western part of the United Kingdom is caused by increase of precipitation amounts. The same reason refer to the increase of cadmium deposition in central part of Kazakhstan, Slovakia, southern part of Greece and Norway, and the central part of Sweden.

35

Fig. 2.20. Total annual deposition of cadmium in Europe and Central Asia in 2008

Fig. 2.21. Spatial distribution of cadmium emissions over the extended EMEP domain in 2008

Annual mean calculated and observed concentrations of cadmium in air in 2008 varied between 0.01 – 0.3 ng/m3 over most part of Europe (Fig. 2.22). The highest calculated concentrations were associated with regions with significant emission sources (Poland and adjacent regions, the Benelux countries, the Balkans). The lowest concentration levels were noted for the Scandinavian countries, the north-eastern part of Russia and the Arctic region. Monitoring network covered territory of the central, western and northern parts of Europe, while in eastern, south-eastern Europe and Central Asia background monitoring data were not available. Therefore, pollution levels over these areas were assessed only by means of modelling.

Fig. 2.22. Calculated and measured surface concentrations of cadmium in air over Europe and Central Asia in 2008

At 2/3 of measurement stations the calculated concentrations of cadmium matched the observed values with bias within ±50% (Fig. 2.23). Most of stations in Austria, the Czech Republic, Germany, France, the United Kingdom, Slovakia, Belgium, Cyprus complied with this criterion. Concentrations in air observed at Slovenian, Finnish, Estonia, Latvian and most of Norwegian stations were underestimated by the model. Significant overestimation (about 7 times) of the observed concentration was noted for Hungarian station HU2. Concentrations measured at HU2 were of similar order of magnitude as those monitored at remote stations such as FI36 (Matarova), GB91 (Banchory) or NO42 (Spitsbergen), and much lower than those measured in neighbouring countries (e.g., the Czech Republic, Slovakia, Germany). Probably, very low measured concentrations can be explained by measurement problems. 0.4 Cd concentrations in air, ng/m

3

Observed Modelled

0.3

0.2

0.1

AT2 AT48 AT5 BE14 CY2 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 EE9 ES8 ES9 FI17 FI36 FI37 FR13 FR9 GB13 GB17 GB91 HU2 IS91 LT15 LV10 LV16 NL10 NL8 NL9 NO1 NO42 PL5 SE14 SI8 SK2 SK4 SK6 SK7

0

Fig. 2.23. Comparison of modelled and measured annual mean concentrations of cadmium in air in 2008 An attempt to link peaks of observed and modelled concentrations of cadmium with emission fields was undertaken for a number of monitoring stations. The analysis for station CZ1 was given here as an example (Fig. 2.24). Back trajectories were drawn for the observed peaks. For some peak periods (e.g., 25th of April, 28th of May, 5th of June, 29th of July) the transport of air masses took place through territory of Poland (Fig. 2.25a). Poland is known for high cadmium emission (Fig. 2.21). Therefore, both modelled and measured values in this period were high compared to other periods. The exception was period of 5th of June. This and previous days were characterized by significant precipitation which lead to scavenging of cadmium in air and hence, to relativity low modelled air concentrations.

36

1.6 1.4 1.2


CZ1

1 0.8 0.6 0.4 0.2 0 11-Jan 21-Jan 31-Jan 10-Feb 20-Feb 03-Mar 13-Mar 23-Mar 02-Apr 12-Apr 22-Apr 02-May 12-May 22-May 01-Jun 11-Jun 21-Jun 03-Jul 15-Jul 25-Jul 04-Aug 14-Aug 24-Aug 13-Oct 23-Oct 02-Nov 18-Nov 02-Dec 12-Dec

Cd concentrations in air, ng/m

3

However, some extremely high peaks could hardly be explained by transport from Poland. For example, in 6th of November (Fig. 2.25b) air were masses arriving to the station passed over Austria and southern part of Europe (Italy, Croatia, and other countries). The agreement between modelled and observed concentrations of cadmium at Austrian stations is satisfactory, so it is hardly possible that emissions in Austria or other countries in this region were so drastically underestimated. Most likely, local sources significantly contributed to levels measured at this station. These situations can be analysed in more detail in the framework of the case study in cooperation with national experts and applying modelling at finer resolution (e.g., 10x10 km2).

Fig. 2.24. Modelled and observed air concentrations of cadmium at Czech station Svratouch (CZ1) in 2008

a

b 2

Emissions, g/km /day

Trajectories

Fig. 2.25. Back trajectories crossing grid cells for the Czech station CZ1 for the period of 29th of July (a) and 6th of November (b) and overall (anthropogenic emission and wind re-suspension) of atmospheric input of cadmium Pattern of cadmium wet deposition was distributed across Europe similar to that of concentrations in air: regions with higher emissions were characterized by higher deposition levels. Since measured deposition fluxes did not always represent entire year both modelled and monitored deposition were expressed as daily mean sums (Fig. 2.26). Relatively high measured and modelled wet deposition fluxes (0.05 – 0.2 g/km2/day) were noted in or nearby regions known for high emissions: in Poland, Slovakia, Belgium, the Netherlands. Relatively small measured and calculated levels were indicated for the United Kingdom (0.2 – 0.4 g/km2/day), Germany and Denmark (0.3 – 0.8 g/km2/day). However, over a number of regions, e.g., Scandinavia, Spain, Hungary and Baltic countries, calculated levels were lower than the observed ones.

37

0.2 Observed 0.15

Modelled

0.1 0.05 0 BE14 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 DK20 DK22 DK31 DK5 DK8 EE11 EE9 ES8 ES9 FI17 FI22 FI36 FI53 FI8 FI92 FI93 FR90 GB13 GB17 GB6 GB91 HU2 IS90 IS91 LT15 LV10 LV16 NL9 NL91 NO1 NO39 NO55 NO56 PL4 PL5 SE51 SE97 SI8 SK2 SK4 SK6 SK7

Cd concentration in precip, μg/L

Comparison of modelled and observed cadmium concentrations in precipitation and wet deposition fluxes is presented in Fig. 2.27 and 2.28. On average, the modelled concentrations and deposition were more than twice lower than the observed ones. For around half of stations calculated concentrations or fluxes differed from the observed values by less then 50%. For most of stations in Germany, the Netherlands, Denmark, Poland, and the United Kingdom and for some Slovak stations the difference lay within ±50% limits. Large discrepancies between modelled and measured values were obtained for stations in the Fig. 2.26. Calculated and measured daily sums Czech Republic, Spain, Latvia, Estonia, Finland, of wet deposition fluxes of cadmium over Europe Norway, and for some stations in Slovakia. The and Central Asia in 2008 discrepancies could be caused by uncertainties of the model (e.g., spatial resolution, parameterizations), measurement problems (e.g., representativeness of station’s location, contamination of samples) or uncertainties of emission data (e.g., completeness, influence of local sources).

120 Observed

2

Cd wet deposition flux, g/km /y

Fig. 2.27. Modelled and measured annual mean concentrations of cadmium in precipitation at individual stations in 2008

100

Modelled

80 60 40 20 BE14 CZ1 CZ3 DE1 DE2 DE3 DE7 DE8 DE9 DK20 DK22 DK31 DK5 DK8 EE11 EE9 ES8 ES9 FI17 FI22 FI36 FI53 FI8 FI92 FI93 FR90 GB13 GB17 GB6 GB91 HU2 IS90 IS91 LT15 LV10 LV16 NL9 NL91 NO1 NO39 NO55 NO56 PL4 PL5 SE51 SE97 SI8 SK2 SK4 SK6 SK7

0

Fig. 2.28. Modelled and measured annual wet deposition flux of cadmium at individual stations in 2008y More detailed analysis of the comparison of measured and modeled cadmium wet deposition fluxes was performed for the Czech station Svratouch (CZ1). In most periods of measurements the observed wet deposition fluxes exceeded the calculated ones (Fig. 2.29). As a rule, the transport of air masses in these periods took place from the western or south-western direction. For example, in the period from 15th to 21st of January the model underestimated measured values by an order of magnitude. Air masses during this period were arriving to the station from territories of Germany, Austria, Italy (Fig.

38

2.30a). The highest contribution to the deposition in this period is made by national sources (Fig. 2.31a). The main external anthropogenic contributors are Germany and Serbia.

Cd deposition flux, g/km

2

7 6 5


CZ1

4 3 2 1 01-Jan 15-Jan 29-Jan 12-Feb 26-Feb 11-Mar 25-Mar 08-Apr 22-Apr 06-May 20-May 03-Jun 17-Jun 01-Jul 15-Jul 29-Jul 12-Aug 26-Aug 09-Sep 23-Sep 07-Oct 21-Oct 04-Nov 18-Nov 02-Dec 16-Dec 30-Dec

0

Fig. 2.29. Modelled and observed concentrations of lead in precipitation at Czech station Svratouch (CZ1) in 2008 There are some periods when the modelled values match the observed ones with high accuracy, for example, in the period from 15th to 21st of April. In this period the transport of air masses took place from the north (Poland, Germany ) and the south/south-east (Slovakia, Hungary etc) (Fig. 2.30b). The major contribution is made by the sources in Poland and Slovakia (Fig. 2.31b), while national sources had comparatively low effect on the modelled wet deposition fluxes. There were also other periods (e.g., 13th – 19th of May) when the agreement between modelled and measured values was satisfactory and the atmospheric transport from Poland and/or Slovakia too place.

a

b Fig. 2.30. Density of back trajectories crossing grid cells for the Czech station CZ1 (left) and total input of cadmium to the atmosphere (anthropogenic emission and re-suspension) (right) for the periods 15th – 21nd of January (a) and 15th – 21st of April (b)

39

2

50 40 30 20 10

500

Poland Re-susp

400 300 200 100

Apr, 21

Apr, 20

Apr, 19

Apr, 18

Apr, 17

Apr, 16

b

Apr, 15

Jan, 21

Jan, 20

Jan, 19

Jan, 18

Jan, 17

Jan, 16

0 Jan, 15

0

a

Czech Rep. Slovakia Other

600

2

Wet dep. of Cd, mg/km /day

60

Germany Re-susp

Wet dep. of Cd, mg/km /day

Czech Rep. Serbia Other

70

Fig. 2.31. Contributions of various sources to modelled concentrations of lead in air at station Svratouch (CZ1) in the periods 15th – 21st of January (a) and 15th – 21st of April (b)

The analysis of individual periods may indicate that the emission data in countries located in the west from the Czech Republic can be underestimated. However, the discrepancies of an order of magnitude imply similar order of the underestimation, which seems unlikely. Besides, in some periods (e.g., 9th – 15th of September) the transport of air masses too place through territory of Poland, but the measured fluxes were nevertheless underestimated by more than 10 times. Probably, this (and other) discrepancies are explained, at least partly, by the uncertainties of measurements or by the influence of local sources. The analysis of this situation should be carried out jointly by MSC-E, CCC and the national experts.

MERCURY Annual total deposition of mercury ranged from 7 to 20 g/km2/y over most part of the EMEP countries (Fig. 2.32). In regions with significant anthropogenic emissions (Fig. 2.33 the deposition exceeded 20 g/km2/y (Poland, the northern part of Italy, the Balkans, the central parts of Russia etc). The lowest deposition was simulated for territories of the northern part of Russia and Scandinavia and over the arid regions of Central Asia.

Fig. 2.32. Total annual deposition of mercury in Europe and Central Asia in 2008

Fig. 2.33. Total annual deposition of mercury in Europe and Central Asia in 2008

40

Over the most countries the country-averaged deposition between 2007 and 2008 changed within ±15%, which was consistent with the changes of deposition caused by annual variability meteorological conditions. For example, the increase of precipitation amount between 2007 and 2008 over the most part of Scandinavia, north-west of Italy and the central and north-western parts of Russia resulted in some increase of total deposition in these regions. Rise of precipitation in Spain was compensated by simultaneous decline of emissions, which finally led to minor changes in deposition in this country. In some countries the deposition changes were more significant. For example, deposition to Denmark declined by almost 20%, to Cyprus – by 45%. The reason for this was the fact that the emissions in 2008 reported by these countries were substantially lower than those in 2007. Similar reason led to decrease of total deposition of mercury in some other countries, e.g., Germany, France, or Norway. In Slovakia and Romania the deposition increased by about 25% and 35%, respectively, because of the increase of emissions. Spatial distribution of mercury concentrations in air was smoother compared to that of lead and cadmium (Fig. 2.34). It is explained by the fact that mercury life time in the atmosphere is about one year. This time is sufficient for almost complete mixing over the globe. It implies that mercury is a global pollutant. Over major part of Europe and Central Asia annual mean modelled and observed concentrations of mercury in air varied between 1.4 to 1.7 ng/m3. In some regions of Europe known for significant mercury emission sources (southern Poland, north of Italy, Hungary, Romania) the concentrations exceeded 1.7 ng/m3. Fig. 2.34. Calculated and measured surface Measurements of elemental or total gaseous mercury concentrations of mercury in air over Europe in air in 2008 were available from 12 stations. The and Central Asia in 2008 concentrations measured at Danish station DK10, British station GB17 and Irish station IE31 were not used in the analysis of annual mean values because their measurement period covers less than one half of a year. Concentrations measured at Polish station PL5 and Czech station CZ3 exhibit very sharp oscillations during the year often decreasing to unrealistically low values, and thus look doubtful. This might be caused by measurement problems because monitoring of mercury is very hard task. The data from the rest eight stations were used in the assessment. Difference between mean annual modelled and observed concentrations of mercury lay within ±15% for all stations (Fig. 2.35). As seen, the best fit between modelled and measured values was obtained for Swedish and Finnish sites. For German and Norwegian stations the model slightly underestimated the observed concentrations.

Hg concentrations in air, ng/m

3

2 1.6 1.2 0.8 Observed

0.4

Modelled 0 DE2

DE8

DE9

FI36

NO1

SE14

NO42

Fig. 2.35. Comparison of modelled and measured annual mean concentrations of mercury in air in 2008 41

Gaseous oxidized and particulate mercury are the forms which readily washed out from the atmosphere. These forms present in the anthropogenic emissions as well as they originate in the atmosphere due to mercury chemical transformations. Therefore, map of mercury wet depositions reflects combined effect of emission distribution, precipitation patterns and regional distribution of oxidizing properties of the atmosphere. Spatial distribution of mercury wet deposition was more inhomogeneous compared to that of concentrations in air (Fig. 2.36), especially in the vicinity of regions with large emissions. Fig. 2.36. Calculated and measured daily sums of wet Over most of the European and Central Asian deposition fluxes of mercury over Europe and Central territory wet deposition fluxes in 2008 ranged 2 Asia in 2008 from 0.01 to 0.05 g/km /day. Observed levels lay in the same range. The highest deposition levels took place in regions known for large emission sources (northern part of Italy, Poland, the Balkan region). The lowest deposition fluxes were noted in the Arctic region and for and in arid regions of Central Asia. The agreement between calculated and most reliable measured mean annual concentrations of mercury in precipitation and wet deposition fluxes was within ±30% for majority of stations and for all stations within ±50% (Fig. 2.37, 2.38). The exception is British site GB91, where the calculated concentrations and fluxes exceeded the observed ones by 2 – 2.5 times.

Hg conc. in precpitation, ng/L

12 10 8 6 4 Observed

2

Modelled

0 BE14 DE1

DE2

DE3

DE8

DE9

ES8

FI96 GB13 GB91 NL91 NO1 SE14

2

Hg wet deposition flux, g/km /y

Fig. 2.37. Modelled and measured annual mean concentrations of mercury in precipitation at individual stations in 2008 14 12 10 8 6 4

Observed

2

Modelled

0 BE14 DE1

DE2

DE3

DE8

DE9

ES8

FI96 GB13 GB91 NL91 NO1 SE14

Fig. 2.38. Modelled and measured annual wet deposition flux of mercury at individual stations in 2008

42

Uncertainties of the model assessment Quality of the air pollution assessment can be characterized on the base of integrated approach which takes into account uncertainties of the model, emission data and measurements. This section summarizes the available estimates of the uncertainties of the emission and monitoring data. These uncertainties are compared with the deviation between modelled and observed concentrations and depositions. Uncertainties of the emission data have been extensively discussed at the recent TFMM meeting held in Larnaca, Cyprus, in May, 2010. It was noted that the uncertainties of the emission estimates can be caused by missing of source categories, insufficient data on activities, current knowledge on emission factors etc. Official information of total emission data uncertainty of HMs in 2008 is available for Denmark, Finland, France, Sweden and the United Kingdom (Table. 2.1). As seen from the table, the uncertainty of country totals of heavy metal emission can reach hundreds of per cents, but typically is about 30 – 60%. Most likely, this is stochastic uncertainty, and it does not include systematic uncertainty of the emissions caused by missing emission sources. Table. 2.1. Uncertainties of heavy metal emission totals for 2008, % Denmark [Nielsen et al., 2010]

Finland [SYKE, 2010]

France [CITEPA, 2010]

Sweden [SEPA, 2010]

United Kingdom [Murrells et al., 2010]

Lead

332

±27

61

15

-30 to +50

Cadmium

195

±29

60

47

-20 to +50

Mercury

142

±25

48

99

-30 to +40

Uncertainties of measurement data also contribute to the discrepancies between modelling results and observations. Intercomparisons of national laboratories involved in the analysis of lead and cadmium measurements sampled at the EMEP stations are carried out regularly. These intercomparisons demonstrate that the accuracy for most of the laboratories is better than ±25% for lead and cadmium in precipitation. Laboratory intercomparisons provide evaluation of the accuracy of analytical methods. Overall measurement accuracy can be estimated by field campaigns. Field comparison of measurements of total gaseous mercury concentrations held in May, 2005, demonstrated that the results of most of the laboratories, participated in the comparison, fall within ±30% range, and for concentrations in precipitation - within ±40% range [Aas, 2006]. Uncertainty of wet deposition of lead and cadmium, estimated on the base of the results of 2006-2007 field campaign, was around 20% [Aas et al., 2009]. Uncertainty of the model assessment of pollution levels can be characterized via comparison of the modelled and measured values, keeping in mind uncertainties of observations mentioned above. Fig. 2.39 demonstrates scatter plots showing deviation of the modelled concentrations in air and wet deposition fluxes from the observed parameters. As seen from Fig. 2.39, for most of stations the deviation of the model value from the measured one is within a factor of two. However, there are some stations for which the discrepancies are significant. For example, the model underestimates wet deposition of lead and cadmium by a factor of three or more at Spanish sites, and some Scandinavian stations. Concentrations of mercury in air are known for low spatial variability, which is of the same order of magnitude as their measurement uncertainties. That is why all modelled and measured concentrations are grouped within a range of 1.4 – 1.8 ng/m3.

43

10

50

2

Model, kg/km /y

Model, ng/m

3

10

1

LV10 ES8

FI37 DK10 FI36

0.1

ES9

NO56

0.1

IS91

0.1

SK2

1

FI36 NO55

1

10

0.1

50

1

10 2

3

Observed, kg/km /y

Observed, ng/m

a 1

2

Model, ng/m

3

Model, g/km /y

100

0.1

ES9 EE9

0.01

LV16

ES8 ES11

10

FR90 ES9

IS91 NO1 FI37

FI53

FI36

1

0.01

NO56 FI22

0.1

1

1

10

3

100 2

Observed, ng/m

Observed, g/km /y

b 50

3

2

Model, ng/m

3

Model, g/m /y

2

1

0.5 0.5

10

2 1 Observed, ng/m

2

3

GB91

2

10

50 2

3

Observed, g/km /y

c Fig. 2.39. Modelled vs. observed concentrations of lead (a), cadmium (b) and mercury (c) in air (left) and wet deposition fluxes (right). Solid red line depicts 1:1 ratio; dashed lines: deviation within factor 2 (red) and factor of 3(green) Table 2.2 summarizes statistical metrics which characterize the model uncertainty. They include Root Mean Square Error (RMSE), correlation coefficient, mean normalized bias (MNB) and a fraction of stations for which the ratio between modelled and measured value falls within a certain factor. For example, F2 relates to two-fold deviation, F3 – three-fold deviation etc. As seen, RMSE for air concentrations and deposition fluxes of heavy metals is around 50%. Keeping in mind uncertainties of the emissions (Table 2.1) and measurements data, this result seems satisfactory. At around 70% of stations modelled and observed lead and cadmium concentrations in air and wet deposition fluxes of lead agree within a factor of two. For wet deposition of cadmium this fraction is around 50%. The MNB indicates that the model has some tendency to underestimate the observed lead and cadmium values. However, the extent of the underestimation is comparable with the emission and measurement uncertainties. Correlation coefficients for all parameters are higher 0.5 44

except for mercury in air. Low correlation for this species is explained by low spatial variability of elemental mercury comparable with the uncertainties of its measurements. It is necessary to note that these statistical parameters are strongly affected by high discrepancies between modelled and values obtained at few stations. Relatively high observed values at these stations can be caused by measurement errors, influence of local emission sources, peculiarities of geographical location of the stations etc. and should be investigated separately in cooperation with CCC and national experts. These investigations could be undertaken in the framework of the EMEP case study project, focused in in-depth analysis of pollution levels in individual countries. Table 2.2. Statistical parameters of the model-to-observation comparison for concentrations in air and wet deposition fluxes Parameter RMSE*, % F2, % F3, % Rcorr MNB**, %

*

RMSE =

1 N

Pb, in air 47 70 89 0.88 -31

⎛M −O ∑i ⎜⎜ i O i ⎝ i N

2

⎞ ⎟⎟ ⋅ 100% ⎠

Pb, wet dep. 45 75 86 0.58 -39

Cd, in air 56 69 85 0.68 -25

Cd, wet dep 54 47 82 0.62 -55

Hg in air 10 100 100 0.2 -7

Hg, wet dep 49 92 100 0.62 1

** MNB = (M − O ) ⋅ 100% O

th M i , Oi –modelled and observed values at i station. M , O - averaged modelled and observed

values

2.2. Preparation of meteorological data for the case study on heavy metal pollution assessment EMEP case study on heavy metal pollution assessment was initiated by MSC-E and EMEP/TFMM [ECE/EB.AIR/GE.1.2009/2]. The main objective of the case study is to improve assessment of pollution levels in the EMEP domain on the base of complex analysis of factors affecting quality of the assessment including emissions, measurements, and modelling, in individual countries. In the framework of this activity it is planned to carry out modelling of pollution levels with higher spatial resolution (e.g., 5x5 km or 10x10 km2) for individual countries. Six countries-volunteers (the Czech Republic, the Netherlands, Croatia, Italy, Spain and Slovakia) have expressed their wish to take part in the case study by now. These countries in cooperation with MSC-E have started the activity in accordance with the programme of the case study [http://tarantula.nilu.no/projects/ccc/tfmm/index.html]. State of the art is presented in [Ilyin et al., 2010] in detail. In this subsection the progress in the field of preparation of input meteorological data for modelling is considered. This activity is part of work package #3 of the programme of the case study. The MM5 [Grell et. al., 1996] model is currently employed as a meteorological pre-processor in regional scale at MSC-E for operational annual calculations within the EMEP region. The same model was used for pilot simulations with high spatial resolution to provide the consistency of meteorological datasets. Model configuration for local-scale calculations was chosen as close to regional-scale one as possible.

45

The methodology of the preparation of meteorological data for air-quality modelling adopted for the case study at MSC-E presupposes three phases: 1. choice of the domain and preparation of the gridded input data for meteorological pre-processor (latitudes and longitudes, map scale factors, topography, land cover and so on); 2. numerical modelling of meteorological processes in itself; 3. analysis of the results. The first (preliminary) phase concerning with domain set up and gridded terrestrial data production has been completed for three of the six countries involved in the case study: the Czech Republic, Croatia and the Netherlands. TERRAIN program (a part of the MM5 modelling system) was used for these purpose. The location of domains boundaries and topography fields for the mentioned countries are presented in Fig. 2.40. The characteristics of grids are given in Table 2.3. As it can be seen, each domain involves not only the considered country, they are several grid points larger on each side to reduce boundary effects.

Fig. 2.40. Domains chosen for air pollution transport modelling within the case study on HM and topography fields Table 2.3. Characteristics of model grids Domain name Czech Republic

Grid cell size, km 5

Number of grid cells

Czech Republic

10

55 × 55

Croatia

10

70 × 55

Netherlands

5

80 × 100

110 × 110

Annual calculations for 2007 (the second phase) have been performed for Czech Republic (on 5×5 km2 and 10×10 km2 grids) and for Croatia (on 10×10 km2 grid) by now. The same set of the parameterizations of the main physical processes as for usual simulations on 50×50 km2 EMEP grid was used, except for cumulus parameterization: the absence of convective precipitation was supposed for 5×5 km2 grid (Table 2.4). The results of calculation for 2007 have been compared to the corresponding results of MM5 regular simulations with 50 km resolution within the EMEP region and to the surface measurements of air temperature at 2 m, wind speed at 10 m and precipitation amount (http://idn.ceos.org – data from

46

Integrated Surface Hourly Database (ISH) archived at the National Climatic Data Centre (NCDC)) (the third phase). Table 2.4 MM5 model configuration Physical process Cumulus Convection Boundary Layer Explicit Microphysics Land Surface

Resolution, km 50, 10 Kain-Fritsch-2 [Kain and Fritsch, 1993; Kain, 2002] MRF [Hong and Pan, 1996] Reisner [Reisner et al., 1998] Five-Layer [Dudhia, 1996]

5 None MRF2 Reisner3 Five-Layer4

The change of spatial resolution allows the refinement of the fields of meteorological parameters taking into account local-scale peculiarities of terrain (Fig. 2.41, 2.42). There are clear correlation between topography and meteorological parameters: in mountain regions of the Czech Republic and Croatia air temperature are lower and precipitation amount are higher than those in plane (compare Fig. 2, 3 on the one hand and Fig. 1 on the other hand).

50×50 km2 grid

10×10 km2 grid

5×5 km2 grid

Fig. 2.41. Spatial distribution of annual mean air temperature at 2m over Czech domain calculated on different model grids

50×50 km2 grid

10×10 km2 grid

Fig. 2.42. Spatial distribution of annual precipitation amount over Croatian domain calculated on different model grids Measurement data with 1-day temporal resolution have been used for the comparison. Only sites with more than 50% of successful measurements have been employed (Table 2.5). A number of statistical

47

indices characterizing the relationship between modelled and measured values of meteorological variables have been used: BIAS, root mean square error (RMSE), linear correlation coefficient (Rcorr), etc. These indicators are calculated on the basis of the sets of timely averaged values of meteorological parameters taken at different points (spatial analysis) and on the basis of the time series at fixed points (temporal analysis). The results of statistical analysis are presented in Tables 2.6-2.11 Number of stations with more than 50% of successful measurements in 2007 within chosen

Table 2.5. domains Domain name Czech Republic Croatia

Table 2.6. Grid cell size, km

Air temperature at 2 m 127 127

Parameter Wind speed at 10 m 125 123

Precipitation amount 76 69

Statistical indices for Czech domain: air temperature at 2m Spatial indices (Based on yearly averaged data)

Temporal indices (Averaged on stations) RMSE, degrees Rcorr Monthly Daily Monthly Daily

Bias, degrees

RMSE, degrees

Rcorr

50 × 50

-0.35

1.60

0.57

1.61

2.34

1.00

0.97

10 × 10

-0.06

1.16

0.80

1.28

1.95

1.00

0.98

5×5

-0.10

0.95

0.87

1.11

1.83

1.00

0.98


Statistical indices for Czech domain: wind speed at 10m Spatial indices (Based on yearly averaged data)

Temporal indices (Averaged on stations) RMSE, m/s Rcorr Monthly Daily Monthly Daily

Bias, m/s

RMSE, m/s

Rcorr

50 × 50

0.49

1.61

0.35

1.31

1.79

0.75

0.72

10 × 10

0.47

1.59

0.35

1.26

1.72

0.79

0.77

5×5

0.23

1.41

0.52

1.07

1.57

0.82

0.77


Statistical indices for Czech domain: precipitation amount Spatial indices (Based on yearly averaged data) Bias, mm

RMSE, mm

Rcorr

50 × 50

17.3

276.1

0.66

10 × 10

-161.9

272.4

5×5

-89.3

193.9


Temporal indices (Averaged on stations) RMSE, mm Rcorr Monthly Daily Monthly Daily 37.9

6.5

0.73

0.37

0.80

37.4

6.6

0.71

0.35

0.80

37.2

7.5

0.71

0.31

Statistical indices for Croatian domain: air temperature at 2m Spatial indices (Based on yearly averaged data)

Temporal indices (Averaged on stations) RMSE, degrees Rcorr Monthly Daily Monthly Daily

Bias, degrees

RMSE, degrees

Rcorr

50 × 50

-1.08

2.33

0.75

1.45

1.67

1.00

0.99

10 × 10

-0.45

1.47

0.89

1.23

1.47

1.00

0.99

48


Statistical indices for Croatian domain: wind speed at 10m Spatial indices (Based on yearly averaged data)

Temporal indices (Averaged on stations) RMSE, m/s Rcorr Monthly Daily Monthly Daily

Bias, m/s

RMSE, m/s

Rcorr

50 × 50

1.13

1.61

0.37

1.22

1.42

0.66

0.76

10 × 10

0.81

1.34

0.52

1.09

1.31

0.78

0.81


Statistical indices for Croatian domain: precipitation amount Spatial indices (Based on yearly averaged data) Bias, mm

RMSE, mm

Temporal indices (Averaged on stations) RMSE, mm Rcorr Monthly Daily Monthly Daily

Rcorr

50 × 50

59.0

277.3

0.47

41.9

7.4

0.66

0.37

10 × 10

-15.0

250.8

0.69

42.7

7.7

0.66

0.40

The following conclusions can be done on the basis of statistical analysis: 1. The growth of spatial resolution results in the improvement of statistical indicators in general. This is true both for Czech and for Croatian domains. 2. All statistics relating to air temperature at 2 m improve substantially. For example, spatial RMSE based on yearly mean values falls from 1.60 degrees to 0.95 degrees with the change of grid resolution within Czech domain from 50 km to 5 km (Table 2.6). At the same time, spatial correlation coefficient increases from 0.57 to 0.87. 3. Statistical indices for wind speed become better too. Thus, temporal RMSE based on monthly values decreases from 1.22 m/s to 1.09 m/s with the change of grid resolution within Croatian domain from 50 km to 10 km (Table 2.10). The corresponding temporal linear correlation coefficient grows from 0.66 to 0.78. 4. Statistics for precipitation amount is ambiguous. Spatial RMSE and Rcorr improve with the increase of grid resolution, but temporal indices do not change essentially. BIAS of annual precipitation values is rather sensitive to grid cell size. But the regularities of its behavior are not clear now. Thus, the change of the grid resolution in Czech domain from 50 to 10 km results in drop of BIAS from 17.3 to -161.9 mm/y (Table 2.8). But the further decrease of cell size down to 5 km leads to growth of BIAS to -89.3 mm/y. At the same time, annual mean station averaged precipitation amount based on the results of the calculation with 10 km resolution in the Croatian domain corresponds to the measurements better than that based on calculation on the 50 km grid (Table 2.11). In addition to integral statistical indicators presented above several examples of particular aspects of comparison of modelled and observed meteorological variables related to individual stations are given below. Thus, the difference between calculated and measured precipitation amount at meteorological stations within Croatian domain is shown in Fig. 2.43. It is seen that the model overestimates precipitation amount in mountain regions of Croatia and Bosnia and Herzegovina on 50 km grid. The increase of spatial resolution results in the improvement of the correspondence between modelling and monitoring data. Annual accumulated precipitation values calculated on 10 km grid differ from those measured no more than 10% at most of meteorological stations.

49

a

b Fig. 2.43. Relative difference between calculated (a - 50 km grid, b - 10 km grid) and observed annual precipitation at meteorological stations within Croatian domain in 2007 (% of measured values)

Fig. 2.44 illustrates the influence of model grid spatial resolution on calculated values of air temperature at 2 m for mountain regions. The decrease of grid sell size makes the correspondence between calculations and measurements at two Czech high-altitude meteorological stations better for all the months of 2007. At the same time, the character of seasonal variations of temperature is successfully reproduced by the model at any resolution.

114570

116430 20 Air temperature, degrees C

114570

Air temperature, degrees C

20 116430

15 10 5 0

Measurements Calculations - 50 km grid Calculations - 10 km grid Calculations - 5 km grid

15 10 5 0

Measurements Calculations - 50 km grid Calculations - 10 km grid Calculations - 5 km grid

-5

-5 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Jan Feb Mar Apr May Jun

Jul Aug Sep Oct Nov Dec

Fig. 2.44. Calculated and measured monthly mean temperature at 2 m in 2007 at meteorological stations 114570 (a) and 116430 (b) located on the territory of the Czech Republic So, the activity concerned with the preparation of meteorological data for the EMEP case study on heavy metals has been started at MSC-E. Annual data sets for the Czech Republic and Croatia have been already prepared and analyzed. For other participating countries the same work are planned to be done this year in accordance with the programme of the case study.

2.3. Comparison of modelled deposition and heavy metal concentrations in mosses Mosses are often used in monitoring of atmospheric pollution [Harmens and Norris, 2008], in particular, deposition of heavy metals. There are several species of mosses which have no roots and thus obtain all nutrients from the atmosphere. Extensive pan-European moss surveys are carried out every five years. ICP-Vegetation of the Working Group on Effects (WGE) is responsible for coordination of this activity. ICP-Vegetation provided MSC-E of EMEP with the data on concentrations of mosses of heavy metals surveyed in 1990, 1995, 2000 and 2005/2006 in 2008. Concentrations of heavy metals in mosses 50

were compared with atmospheric depositions simulated by EMEP MSCE-HM model [Gusev et al., 2009]. Spatial variability of total deposition and concentrations in mosses in Europe were compared, long-term trends of the deposition and concentrations were analyzed. Besides, correlation coefficients between the deposition and concentrations were calculated for each country for each survey. Concentrations in mosses and deposition fluxes cannot be compared directly. Nevertheless, comparability between these two parameters can be characterized by spatial correlation coefficient. According to the results presented in [Gusev et al., 2009] in some countries, for example, in Finland, Sweden or the Czech Republic, the correlation coefficient was relatively high (0.7 – 0.8). However, in many other countries it was relatively low or even negative. This year more detailed analysis of spatial distribution of pollution levels simulated by the model and monitored by moss surveys was performed. The aim of this work is to consider factors which may affect the correlation coefficient. In previous steps of the analysis [Gusev et al., 2009] total deposition fluxes averaged over three years prior to moss surveys were used. These deposition data will be considered as “base-case” results. Base-case spatial correlation coefficients will be compared with those calculated in the numerical experiments aimed at consideration of factors affecting the comparison of deposition and concentration in mosses. Four factors were considered. First of all, the effect of type of deposition was analyzed. Concentrations in mosses were compared separately with modelled wet and dry deposition fluxes. Second factor considered was period of averaging of deposition. Furthermore, the effect of modelled wind re-suspension on the comparison was evaluated. Finally, the influence of variability of measured moss concentrations within EMEP grid cells and density of sampling network was characterized. The correlation between concentrations in mosses (MC) and different types of deposition varies largely among countries (Fig. 2.45). For example, in case of lead in 2005, it is possible to single out several groups of countries. First of all, there are countries where the correlations between moss concentrations and total (TD), wet (WD) and dry (DD) deposition fluxes are similar and relatively high. These are Poland, Finland, the Czech Republic and some others. In some countries, e.g., Croatia, Lithuania, Latvia, correlation of MC vs. WD is significantly higher than that MC vs. DD or MC vs. TD. In the third group of countries correlation coefficient MC vs. DD is higher than that of MC vs. WD or MC vs. TD. These are Austria, Germany, the United Kingdom.

MC vs. TD

MC vs. WD

MC vs. DD

Pb, 2005

U.K.

Ukraine

Turkey

Switzerland

Spain

Sweden

Slovenia

Serbia

Slovakia

Poland

Norway

Latvia

Lithuania

Iceland

Germany

FYR_Macedonia

France

Finland

Croatia

Czech_Republic

Bulgaria

Belgium

0.4 0.2 0 -0.2 -0.4 -0.6 Austria

Correlation coeff.

1 0.8 0.6

Fig. 2.45. Spatial correlation coefficients between concentrations of Pb in mosses (MC) sampled in 2005 and total deposition (TD, ‘base case’), wet deposition (WD) and dry deposition (DD)

51

4

8 6 4

4

0.7 - 1

0.5 - 0.7

0.3 - 0.5

< 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5

< 0.3

0

0.7 - 1

0

0.5 - 0.7

0 0.3 - 0.5

2

0 < 0.3

2

Cd, 2005

Range of correlation coef.

Cd, 2000

Cd, 1995

0.7 - 1

6

2


0.5 - 0.7

8

2


Total Wet Dry

10

0.7 - 1

6

10

12

0.5 - 0.7

8

Pb, 1990 Total Wet Dry

12

0.3 - 0.5

< 0.3

0.7 - 1

< 0.3


0.3 - 0.5

4

0

# of countries

6

10

2

Pb, 1995 # of countries

8

3 1

14

Total Wet Dry

12 # of countries

10

14

4


Pb, 2000 Total Wet Dry

12

0.5 - 0.7


Pb, 2005 14

0.3 - 0.5

< 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5

0

5

< 0.3

2

6 # of countries

4

Total Wet Dry

7

0.7 - 1

6

Total Wet Dry

10 9 8 7 6 5 4 3 2 1 0 0.5 - 0.7

8


# of countries

Total Wet Dry

10

0.3 - 0.5

12

# of countries

Total Wet Dry

# of countries

10 9 8 7 6 5 4 3 2 1 0 < 0.3

# of countries

Frequency distribution of number of countries corresponding to correlation coefficient ranges is presented in Fig. 2.46. As seen, number of countries with significant correlation coefficient (more than 0.5) is typically higher when concentration in mosses compared with total deposition rather than with wet or dry deposition separately. General conclusion from these results is that for comparison of deposition and concentrations in mosses both wet and dry deposition fluxes are important.


Cd, 1990

Fig. 2.46. Frequency distribution of spatial correlation coefficients in countries between concentrations of Pb and Cd in mosses and total, wet and dry deposition When sampling of mosses is carried out green parts of mosses are collected. It is assumed that their age is about 3 years [Harmens and Norris, 2008]. Therefore, the concentration in mosses represents accumulated value for these 3 years, and deposition fluxes used in this comparison are also averaged over there years. In order to test the importance of the period of averaging on the comparison results, deposition fluxes averaged over there years (‘base case’) and over one year prior to moss survey were compared with heavy metal concentrations in mosses. Change of averaging period of modeled deposition had minor effect on the correlation between total deposition and concentrations in mosses (Fig. 2.47). The main reason of this is small changes in annual deposition patterns from one year to another. According to the model calculations, significant contribution to deposition of lead and cadmium is made by wind re-suspension. For example, in 2007 a number of countries contribution of resuspension of lead to deposition exceeded 50% in about half of European and Central Asian countries [Ilyin et al., 2009]. However, the parameterization of wind re-suspension of metals contains significant uncertainties. In order to test the effect of modelled wind re-suspension on the correlation between deposition and concentrations in mosses (MC), calculations of metal pollution caused by only anthropogenic sources (AS) and only by wind re-suspension (RS) were carried out.

52

Pb, 2000

0.6 0.4 0.2 0 -0.2

Russia_(Central) Russia_(NorthWest)

Ukraine

United_Kingdom

Switzerland

Slovakia

Poland

Portugal

Norway

Lithuania

FYR_Macedonia

Italy

Latvia

Iceland

Hungary

France

Germany

Finland

Estonia

Bulgaria

Czech_Republic

Austria

Belgium

-0.6

Spain

3-year averaging 1 year averaging

-0.4

Sweden

Correlation coeff.

1 0.8

Fig. 2.47. Spatial correlation coefficients between concentrations of Pb in mosses (MC) sampled in 2000 and total deposition fluxes averaged over 3 years and 1 year Example of correlation coefficients between concentrations in mosses and total deposition of lead for 2005 survey is demonstrated in Fig. 2.48. As seen from this figure, inclusion of wind re-suspension may have either positive or negative effect on the correlation in different countries. For example, in Austria, Lithuania and the United Kingdom the correlation MC vs. RS is significantly higher than the correlation MC vs. AS. In most cases the countries the correlation between deposition from all sources (‘base case’) and MC is similar to that of MC vs. RS and MC vs. AS. In means that both anthropogenic sources and re-suspension are important for the evaluation of pollution levels of heavy metals. This idea is also confirmed by the analysis of frequency distribution of correlation coefficients for the lead and cadmium for surveys of 1990, 1995, 2000 and 2005 (Fig. 2.49).

1 0.8 0.6 0.4 0.2 0 -0.2 -0.4

Re-suspension

U.K.

Ukraine

Turkey

Sweden

Switzerland

Spain

Slovakia

Slovenia

Serbia

Poland

Norway

Lithuania

Latvia

Iceland

Germany

France

Finland

CzechRepublic

Croatia

Anthrop.

The_FYR_Macedonia

All sources

Bulgaria

Austria

Belgium

Pb, 2005

Fig. 2.48. Spatial correlation coefficients between concentrations of Pb in mosses (MC) sampled in 2005 and total deposition based on all emission sources (All), only anthropogenic sources (Anthrop) and only on wind resuspension (Re-susp) Final set of experiments was aimed at the evaluation of the effect of spatial variability of moss measurements within EMEP grid cells on the correlation coefficients. Density of moss measurements varies largely across Europe. It means that number of moss sampling plot falling in each the EMEP grid cells differs among the grid cells. In most of the EMEP grid cells there are one, two or three sampling plots (Fig. 2.50). In the quarter of the grid cells number of moss sampling plots equals or exceeds five.

53

Cd, 2005

4

0.7 - 1

0.5 - 0.7

6 4 2

Cd, 1995

0.7 - 1

0.5 - 0.7

< 0.3

0.7 - 1

0.5 - 0.7


Cd, 2000

All Anthrop. Resusp.

8

0 < 0.3

0.7 - 1


0.3 - 0.5

< 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5

8

0 0.5 - 0.7

0.7 - 1

0 0.5 - 0.7

0 0.3 - 0.5

2

12

10 # of countries

4

2


# of countries

6

0.3 - 0.5

4


Pb, 1990 All Anthrop. Resusp.

16

8

< 0.3

6

10

4 3 2 1 0

Pb, 1995 20


12 # of countries

8

< 0.3

# of countries

10

14

6 5


Pb, 2000 All Anthrop. Resusp.

12

< 0.3


Pb, 2005 14

0

0.3 - 0.5


4 2

< 0.3

0.7 - 1

0 0.5 - 0.7

2

0 0.3 - 0.5

2

6


0.3 - 0.5

4

8

8 7 # of countries

6

# of countries

8


10

0.7 - 1

4


0.5 - 0.7

6

< 0.3

# of countries

8

10

0.3 - 0.5


10

# of countries

12


Cd, 1990

Fig. 2.49. Frequency distribution of spatial correlation coefficients in countries between concentrations of Pb and Cd in mosses total deposition based on all sources, anthropogenic sources and re-suspension

25 Total: 1534 grid cells

% of gridcells

20 15 10 5

1 2 3 4 5 6 7 8 9 10 > 10

0 Number of samples per grid cell

a b Fig. 2.50. Number of moss sampling plots in the EMEP grid cells (a) and and frequency distribution of this number (b) Concentrations in mosses within EMEP grid cells can vary significantly. Fig. 2.51a depicts ratio of maximum to minimum measured concentrations in mosses within EMEP grid cells. High variability is noted for Switzerland, Germany, Slovakia, and the Balkan region. Relatively small ration in France and the United Kingdom is explained by the fact that sampling density in these countries is 1 – 2 plots per grid cell. In more than a half of grid cells the ratio is higher than 1.5, and in 1/3 of grid cells – more than 2 (Fig. 2.51b). When concentrations in mosses are compared with modelled deposition fluxes, their values are averaged over the model grid cell. If a number of plots within a grid cell is small (e.g., 1 or 2) it is not possible to be sure if they characterize higher or lower or medium pollution levels within a grid cell.

54

Therefore, modeled depositions were compared with grid cells where three or more sampling plots are located. The results of this exercise were compared with the base case.

30 Total: 1534 grid cells

% of gridcells

25 20 15 10 5

> 10

5 - 10

3-5

2-3

1.5 - 2

1

1 - 1.5

0

Max / Min ratio

a b Fig. 2.51. Ratio of minimum-to-maximum concentration of lead in mosses sampled in 2005 (a) and frequency distribution of the ratio (b) Correlation between total deposition and concentrations in mosses is better for areas with higher sampling density. For example, in case of lead surveyed in 2005 significant increase of correlation coefficient is noted for France, the United Kingdom, the Ukraine and other countries (Fig. 2.52). Frequency distribution of number of countries falling in ranges of correlation coefficients is shown in Fig. 2.53. As seen, the tendency of increasing of the correlation coefficient as density of sampling increases is noted for all four surveys both for lead and for cadmium. It is important to note that in diagrams shown in Fig. 2.53 number of countries related to the case of higher sampling density may be lower than that in base case. The reason is that is some countries number of grid cells with three or more sampling plots is small, and thus calculation of correlation coefficient becomes not correct.

Correlation coeff.

1

Pb, 2005

0.8 0.6 0.4 0.2 0 -0.2 U.K.

Ukraine

Turkey

Switzerland

Sweden

Spain

Slovenia

Slovakia

Serbia

Poland

Norway

Lithuania

FYR_Macedonia

Latvia

Iceland

France

Germany

Finland

Czech_Republic

Croatia

Bulgaria

Austria

-0.4

All grid cells 3 or more samples in a grid cell

Fig. 2.52. Spatial correlation coefficients between concentrations of Pb in mosses sampled in 2005 and total deposition for different sampling density

55

8 6 4

4

Cd, 2005

0.7 - 1

0.5 - 0.7

0.3 - 0.5

< 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5

< 0.3

0

0.7 - 1

0

0.5 - 0.7

2

0 0.3 - 0.5

2

0 < 0.3

2



Cd, 2000

Cd, 1995

0.7 - 1

6

2


0.5 - 0.7

8

0.7 - 1

4

0.3 - 0.5

< 0.3

< 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5

6

10

Base case 3 or more samples

10

0.5 - 0.7

4

8

12

0.3 - 0.5

6

12

< 0.3

8

Pb, 1990


14 # of countries

# of countries

3 or more samples


Pb, 1995


10

2 0

# of countries

12

3 or more samples

4


Pb, 2000

Base case

10

2


Pb, 2005 12

4

0 < 0.3

0.7 - 1

0.5 - 0.7

0.3 - 0.5


# of countries

2 0

< 0.3

0

4

6

Base case

6

0.7 - 1

2

6

0.5 - 0.7

4

8

8 # of countries

6


10

3 or more samples

8 # of countries

# of countries

8

# of countries

Base case

10

0.3 - 0.5


10


Cd, 1990

Fig. 2.53. Frequency distribution of spatial correlation coefficients in countries between concentrations of Pb and Cd in mosses and total deposition if 3 or more moss sampling points falling in EMEP grid cell and in the base case (all grid cells)

56

3. PERSISTENT ORGANIC POLLUTANTS This chapter is aimed at the comparison of calculation results obtained by MSCE-POP model with available measurement data obtained at the EMEP monitoring network. The section is divided into two parts. First part contains general comparison of measured and calculated air concentrations, concentrations in precipitation and deposition fluxes for B[a]P, PCB-153 and γ-HCH. The second part is devoted to the elaboration of complex monitoring/modelling/emission approach based on the usage of back trajectories. This approach is exemplified by the pollutant with maximum available measurement information – B[a]P. At present the analysis is of preliminary character. During next year it is planned to further develop the approach and apply it to the examination of pollution by all considered pollutants.

3.1. Comparison with measurements BENZO[a]PYRENE (B[a]P)

factor 3

0.5 0.4

factor 2

0.3 0.2 0.1

20

factor 3 15

10

5

DE1 EE9 NO1 SI8 CZ3 SE12 DE9 ES8 LV10 SE14 NO42 DE8 PL5 DE3 GB14 LV16 FI96

10 8 6 4

0

0

b

12

2

0

a

Measured Calculated

14

2

0.6

Measured Calculated

25

B[a]P deposition flux, ng/m /day

Measured Calculated

0.7

B[a]P concentration in prec., ng/L

B[a]P air concentrations, ng/m

3

The measurement data of B[a]P concentrations in the atmosphere is available at 17 EMEP measurement sites, namely CZ3, DE1, DE3, DE8, DE9, EE9, ES8, FI96, GB14, LV10, LV16, NO1, NO42, PL5, SE12, SE14 and SI8. The results of the comparison of calculated air concentrations with measurements at these sites (annual averages) are shown at Fig. 3.1a.

CZ3

DE1

DE3

DE8

DE9

PL5

c

FI96

SE12

SE14

Fig. 3.1. Comparison of calculation results with measurements for B[a]P: (a) – air concentrations, ng/m3; (b) – concentrations in precipitation, ng/L; (c) – deposition flux (annual averages), ng/m2/day

Among them 6 sites (CZ3, DE1, DE3, DE8, DE9 and PL5) perform parallel measurements of B[a]P concentrations in precipitation and 3 sites (FI96, SE12 and SE14) – of deposition flux. The existence of simultaneous measurements of air concentrations and concentration in precipitation/deposition flux allows evaluating model parameterization of wet deposition process. First of all, we note that calculations agree with measurements within a factor of two at 10 sites, and within a factor of three – at three sites more. The maximum ratio between calculations and measurements is 4.7. This value is achieved at site FI96, where measured annual mean of B[a]P concentration is as low as 0.005 ng/m3. In general it can be seen that calculated values of air concentrations exceed measurements 1.7 times on the average.

57

The agreement between calculation and measurements of B[a]P concentrations in precipitation is mostly within a factor of three. The only exception is site CZ3, where calculated-to measured ratio is 4.02. Calculations of concentrations in precipitation exceed measurements 3.24 times on the average. Finally, measured and calculated values of deposition flux differ not more than 2.4 times. At that, calculation results are lower than measurements at two sites (FI96 and SE12) and higher – at SE14. The ratio of calculated to measured values of air concentrations, concentrations in precipitation and deposition flux are summarized in Table 3.1. Table 3.1. Calculation-to-measurement ratios for air concentrations, concentrations in precipitation and deposition flux Site CZ3 DE1 DE3 DE8 DE9 PL5 FI96 SE12 SE14

Air concentrations 0.86 0.98 3.28 2.34 0.69 0.41 4.79 1.43 1.87

Calculated-to-measured ratio Concentrations in precipitation 4.02 2.27 2.88 2.75 1.71 0.41 – – –

Deposition flux – – – – – – 0.86 0.75 2.40

It can be found that overestimation of concentrations in precipitation at a number of sites is much stronger than overestimation of air concentrations. In particular, at CZ3, DE1 and DE9 the calculated air concentrations are less than measured ones, whereas calculated concentrations in precipitations exceed measured ones several times. At the rest sites where concentrations in precipitation are measured (DE3, DE8 and PL5) calculated-to-measured ratio for air concentrations and concentrations in precipitation correspond well to each other. On the opposite, calculation-to-measurement ratio for air and deposition flux at SE12 and SE14 correspond well to each other. The difference between calculation-to-measurement ratios at FI96 can be conditioned by higher uncertainty in measurements of B[a]P in low concentration range. In any case, it can be claimed that the comparison does not show overestimation of deposition fluxes by the model. Such a situation (overestimation of concentration in precipitation and reasonable agreement between values of deposition fluxes) can be explained by several reasons. First, it can be caused by the uncertainties of model parameterizations of dry and wet deposition processes. Second, the reason may be the difference of actual temperature and precipitation intensity at measurement sites and the values of these parameters used in the model for the corresponding 50×50 km cells. For further investigation of this phenomenon, simultaneous measurements of concentrations in precipitation, deposition flux and air concentrations at various sites with different locations are strongly desirable. Let us now proceed with the comparison of seasonal variations of B[a]P concentrations obtained from model calculations and measurements. Monthly means of calculated and measured B[a]P air concentrations at all 17 measurement sites are presented by plots in Fig. 3.2.

58

Jan

Apr

59

1.5

1

0.5

0

2

SI8 0.3

SE12

2.5

Measured

1.5

Calculated

1

0.5

0

Fig. 3.2. Comparison of monthly means of calculated and measured air B[a]P concentrations at 17 EMEP measurement sites Nov Dec

NO1

Oct

0.14 0.12

Nov

Calculated Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

GB14

Dec

Measured

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Feb

EE9

Oct

0 Aug

0.1

3

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Feb

Jan

Dec

Nov

Oct

Sep

3


DE3

Sep

0.2

Aug

0.3

0.35

Sep

0.4

Jul

0.5

Jul

Calculated 0.18 0.16

Jun

0

Jun

0.01

Apr

0.02

Apr

0.03

May

0.04 0.5

Apr

Calculated

May

0.05

Dec

Measured

Mar

0

Feb

0.2

0.6

Mar

0.4

Feb

0.6

Jan

1

B[a]P air concentration, ng/m

0.8

Jan

3

0.07


Nov Dec

Calculated

Mar

3

Measured


Nov Dec

Oct

Measured

Jan

Dec

Oct

Aug Sep

0

Feb

3

2.5


Oct Nov

Aug Sep

0.1

Jan

2

Dec

PL5

Oct

0.6

Nov

Aug

LV16

Sep

Jul Aug

0.2

Oct

0 Jul

Jun

0.3

Nov

0.05

Jul

0.4

0.3

Sep

0.1

Jun

May

Apr

Mar

Calculated

Jul

0.15 Jun

Jan Feb

3


Measured

Aug

Calculated Jul

May

Apr May

3

0.6

Jun

Measured Jun

Apr

FI96

Apr

0.2

Mar

1.2

May

SE14 Mar

DE9

Mar

0

1.6

Feb

0.005

Aug

0.01

DE1

Sep

0.015

Jul

Calculated

Jun

Measured 0.8

Jan

0.25

3

0


0.1

May

0.2

May

0.3

Dec

0.4 0.7

Apr

Calculated 0.06

Apr

Calculated

Feb

0

Mar

0.1

Mar

Jan

0.2

Feb

0.3

Jan

0.4 1.4

Oct

Jan

Calculated


Measured

Feb

3

0.6

Nov

Jan

Measured


Dec

0.1 0

0.5

Feb

3

Dec

Oct Nov

0.2

Aug

Measured


Dec

Nov

0.4 0.3

Sep

3

0.025


Dec

Nov

Sep

Aug

Jul

Jun

May

Apr

Mar

Calculated

Jul

Dec

Nov

Jan Feb

Measured

Jun

May

0.02

Nov

ES8

Mar

NO42 Oct

Sep

DE8

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

0.5

Oct

0.5

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan Feb

3

B[a]P air concentrations, ng/m 0.7 0.6 0.5

Oct

Aug

Jul

Jun

LV10

Sep

Aug

Jul

Jun

May

Jan Feb

3


CZ3

Feb

3


May

Apr

Apr

0.6

Mar

0.7

Mar

Jan

3

B[a]P air concentrations, ng/m 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Feb

3

B[a]P air concentrations, ng/m 0.8

Feb

Jan

3

B[a]P air concentrations, ng/m 0.9

Measured

0.25

Calculated

0.15

0.2

0.05

0.1

0

0.7 Measured

0.5

Calculated

0.4

0.3

0.2

0.1 0

0.6 Measured

0.4

Calculated

0.3

0.2

0.1 0

Measured

Calculated

0.08 0.06 0.04 0.1

0.02 0

0.35

Measured

0.25 Calculated

0.15 0.2

0.05 0.1

0

From these plots it is seen that in spite of total overestimation of air concentrations, calculation results well correlate with measurements. Correlation coefficients for all measurement sites are given in Table 3.2. Table 3.2. Correlation coefficients between monthly averages of calculated and measured air concentrations at 17 EMEP sites CZ3 DE1 DE3 DE8 DE9 EE9

0.88 0.56 0.68 0.65 0.77 0.77

ES8 FI96 GB14 LV10 LV16 NO1

-0.44 0.82 0.85 0.73 0.77 0.95

NO42 PL5 SE12 SE14 SI8

0.76 0.86 0.62 0.92 0.86

It is seen that correlation coefficients are 0.6 and higher. The only poor correlation is obtained at the site ES8. To explain this correlation, we remark that beginning from August 13 of 21 measured values of air concentrations at this site were below the detection limit. The plot of raw measurement data in comparison with calculations is presented in Fig. 3.3.

3.4

0.5

1.2


3

Measured Calculated

0.4 0.3 0.2 0.1

14 dec 22 dec

13 dec

27 nov 5 dec

19 nov

3 nov 11 nov

26 oct

9 oct

17 oct 25 oct

23 sep 1 oct

15 sep

6 sep 7 sep

29 aug

5 aug

13 aug 21 aug

9 jun 17 jun

1 jun

23 may 31 may

15 may

29 apr 7 may

21 apr

4 apr

12 apr 13 apr

3 mar

11 mar 19 mar

23 feb 24 feb

15 feb

0

Fig. 3.3. Comparison of raw measurements with calculated B[a]P air concentrations at ES8 It can be seen that if outstanding values of air concentrations (3.4 ng/m3 at August 29 and 1.2 ng/m3 at September 7) and concentrations below the detection limit are ignored, the agreement of measured and calculated data becomes much better. At the same time, the analysis of the plots in Fig. 3.3 leads to the conclusion that the model underestimate seasonal variations of air concentrations at most sites. This can be conditioned by underestimation of emission seasonal variations. So, it can be concluded that: 1. Model calculations show reasonable agreement with measurements. Calculated values of air concentrations agree with measurements within a factor of two at about 60% of measurement sites, and within a factor of three – at about 75% of sites. At the same time calculations exceed measurements 1.7 times on the average.

60

2. Correlation coefficients of calculated monthly averages with measured ones are normally in the range from 0.6 to 0.95. However, seasonal variations of air concentrations are underestimated by the model. 3. Further work is needed for the refinement of model parameterization of dry and wet deposition of B[a]P. More precise analysis of overestimation of air concentrations and their seasonal variations with the help of back trajectory approach will be given below (see Section 3.2).

POLYCHLORINATED BIPHENYLS (PCB-153)

3.5 Measured 3

Calculated

2.5 2 1.5 1 0.5 0

a

PCB-153 deposition flux, ng/m2/day

PCB-153 concentration in prec., ng/L

PCB-163 air concentrations, pg/m 3

Measurement data on PCB-153 concentrations in the ambient air are available at 7 EMEP monitoring sites: CZ3, FI96, GB14, IS91, NO1, NO42 and SE12. The plot of the comparison of measured and calculated air concentrations for the pollutant is given in Fig. 3.4a. 0.25 Measured Calculated

0.2 0.15 0.1 0.05 0

CZ3 FI96 GB14 IS91 NO1 NO42 SE14

b

1.4 1.2

Measured Calculated

1 0.8 0.6 0.4 0.2 0

CZ3 DE1 DE3 DE8 DE9 IS91 NO1

c

FI96

SE12

SE14

Fig. 3.4. Comparison of calculation results with measurements for PCB-153: (a) – air concentrations; (b) – concentrations in precipitation; (c) – deposition fluxes (annual averages) It is seen that calculated values of air concentrations agree with measured ones within a factor of two. The correlation coefficient between calculations and measurements is 0.89. The value of calculationto-measurement ratio is 0.79 on the average, so that slight underestimation of air concentration by the model takes place. The data on PCB-153 concentrations in precipitation is available at EMEP sites CZ3, DE1, DE3, DE8, DE9, IS91 and NO1. The comparison of calculated values of PCB-153 concentrations in precipitation with measurements at these sites is displayed in Fig. 3.4b. Calculation-to-measurement ratio for the considered parameter is 0.35 with correlation coefficient 0.53. Deposition fluxes are measured at EMEP sites FI96, SE12 and SE14. The model calculations agree well with measurements at site FI96 and show higher deposition fluxes than measured ones at SE12 and SE14 (see Fig. 3.4c). In general, the agreement of calculated and measured values of the flux is within an order of magnitude. Possible reasons of overestimation of deposition flux by the model will be discussed later in the course of consideration of seasonal variations of pollution by PCB-153. The consideration of seasonal variations of air concentrations will be performed at the following 6 measurement sites: CZ3, FI96, IS91, NO1, NO42 and SE12. Site GB14 is excluded from the consideration since monthly means of concentrations are not available at the site. Monthly means of calculated and measured PCB-153 air concentrations at the above sites are presented by plots in Fig. 3.5.

61

6 jan 13 jan 20 jan 27 jan 3 feb 10 feb 17 feb 24 feb 2 mar 9 mar 17 mar 24 mar 30 mar 6 apr 13 apr 20 apr 27 apr 4 may 11 may 18 may 25 may 1 jun 8 jun 15 jun 22 jun 29 jun 6 jul 13 jul 20 jul 27 jul 3 aug 10 aug 17 aug 24 aug 31 aug 7 sep 14 sep 21 sep 28 sep 5 oct 12 oct 19 oct 26 oct 2 nov 9 nov 16 nov 23 nov 1 dec 7 dec 14 dec 21 dec 28 dec

3

PCB-153 air concentrations, pg/m

2 jan 9 jan 16 jan 23 jan 30 jan 6 feb 13 feb 20 feb 27 feb 5 mar 12 mar 19 mar 26 mar 2 apr 9 apr 16 apr 23 apr 30 apr 7 may 14 may 21 may 28 may 4 jun 11 jun 18 jun 2 jul 4 jul 9 jul 11 jul 16 jul 23 jul 30 jul 6 aug 13 aug 20 aug 27 aug 3 sep 4 sep 10 sep 17 sep 24 sep 1 oct 8 oct 15 oct 16 oct 22 oct 29 oct 5 nov 12 nov 20 nov 26 nov 3 dec 10 dec 17 dec 24 dec

3

PCB-153 air concentrations, pg/m Jan

1.8

1.6

1.4

2.5

2

62

Jul

Jun

0.6

0.4

0.2

0 Jul

Jun

3

SE14

Fig. 3.7. Comparison of raw measurements with calculated PCB-153 air concentrations at NO42 Nov

Fig. 3.5. Comparison of monthly means of calculated and measured air PCB-153 concentrations at 6 EMEP measurement sites

Correlation coefficients Kcorr between measurements and calculation results are in the range from 0.64 to 0.9 with exception for the sites NO1 (Kcorr = 0.21) and NO42 (Kcorr = 0.19). To explain these two correlations, we consider the plot of raw measurement data in comparison with calculations for these two sites (Figs. 3.6 and 3.7). 2

Measured

Calculated

1.2

0.8 1

0.6

0.4

0.2

0

Fig. 3.6. Comparison of raw measurements with calculated PCB-153 air concentrations at NO1

3

Measured

Calculated

1.5

1

0.5

0 Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

IS91

Dec

0.8

0.6

Oct

1

Sep

Calculated

3.5

Aug

Measured

Jan

0 Feb

0.1

3

3


0.2

May

1.4 3

0.3


Dec

Nov

Oct

Sep

Aug

Jul

Jun

0.4

0.7

Apr

NO42

Nov

Apr May

Calculated

Mar

0

Measured

Jan

0.2

0.6

Feb

0.4

FI96

Dec

0.6 0.5

Oct

0.8

Sep

1

Aug

Calculated 1.2

May

Measured

Mar

0

Apr

Calculated

Mar

Measured

Jan

2

Feb

3


4

Feb

1.6 3

CZ3


Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

5

Jan

Dec

1.2

Nov

NO1

Oct

Sep

Aug

Jul

Jan Feb

3

PCB-153 air concentrations, pg/m 1

Jun

1.4

May

Apr

Mar

Feb

3


6 0.8 Measured

0.5

Calculated

0.4

0.3

0.2

0.1 0

4 Measured

2.5

Calculated

1.5

2

0.5

1

0

For site NO1, it is seen that the discrepancies between model calculations and measurements are found for two periods: from January 23 to March 12 and from April 2 to April 23. For the rest time the model reasonably reproduces PCB-153 air concentrations. For NO42 there exist two episodes where calculations strongly differ from measurement data: June 15 and August 17 and 24. So, the discrepancies between calculations and measurements at these two sites may be explained by the influence of local emission sources. For more detailed analysis of the situation trajectory approach (see Section 3.2 below) should be applied. This is planned to be performed in future. From the consideration of the comparison of calculated and measured monthly means of PCB-153 air concentrations (Fig. 3.5) it can be also concluded that the model underestimates seasonal variations of the pollution by PCB-153. This may be conditioned by the absence of information on emission seasonal variations for the considered pollutant. At present the emissions of PCB-153 are assumed to be distributed homogeneously within the year.

Dec

Oct

Nov

Sep

Jul

Aug

0 Jun

b

1

May

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

0

2

Apr

0.5

3

Mar

1

Measured Calculated

Jan

1.5

Jan

SE14 4

Feb

Calculated

2

5

2

Measured

SE12

Feb

a

PCB-153 deposition flux, ng/m /day

2.5

2

PCB-153 deposition flux, ng/m /day

Let us turn back to the consideration of the discrepancies between calculated and measured PCB-153 deposition flux at sites SE12 and SE14. The plots of monthly averages of deposition fluxes at these sites are shown in Fig. 3.8.

Fig. 3.8. Comparison of measured and calculated PCB-153 deposition fluxes at SE12 and SE14 From these plots it is clear that overestimation of PCB-153 deposition flux takes place mainly in winter months. Hence, further refinement of model parameterization of deposition flux is required. So, it can be concluded that: 1. Model calculations show reasonable agreement with measurements. Annual means of PCB-153 air concentrations agree with measured ones within a factor of two. The correlation coefficient between calculations and measurements is 0.89. At the same time calculated values of annual means of air concentrations are lower than measured ones by about 20% on the average. 2. Correlation coefficients of calculated monthly averages with measured ones are normally in the range from 0.64 to 0.9. Poor correlation between measurements and calculations at NO1 and NO42 may be conditioned by local emission sources. Further analysis of these discrepancies with the help of back trajectory approach is planned in future. 3. Seasonal variations of air concentrations are underestimated by the model. For refinement of model representation of PCB-153 seasonal variation the information on seasonal variations of emissions is desirable. 4. Further work is needed for the refinement of model parameterization of deposition flux of PCB153, especially in the cold periods. For the refinement of model description of PCB-153 depositions more simultaneous measurements of air concentrations, concentrations and deposition flux are required.

63

LINDANE (γ-HCH)

γ-HCH deposition flux, ng/m /day

2.5 Measured

2

NO1

0 IS91

b

0.5 NL91

SE14

SE12

NO1

NO42

FI96

IS91

DE9

0

1

DE9

5

1.5

DE8

10

2

DE3

15

Calculated

CZ3

20

Measured

DE1

Calculated

3 2.5

BE14

γ-HCH concentration in prec., ng/L

Measured

25

CZ3

a

30

DE1

γ-HCH air concentrations, pg/m

3

Measurement data on γ-HCH concentrations in the ambient air are available at 9 EMEP monitoring sites: CZ3, DE1, DE9, FI96, IS91, NO1, NO42, SE12 and SE14. The plot of the comparison of measured and calculated air concentrations for the pollutant is given in Fig. 3.9a.

2

Calculated

1.5 1 0.5 0

c

FI96

SE12

SE14

Fig. 3.9. Comparison of calculation results with measurements for γ-HCH: (a) – air concentrations; (b) – concentrations in precipitation; (c) – deposition fluxes (annual averages) The agreement between calculated and measured annual means of γ-HCH air concentrations is within a factor of three at all sites except for NO42 and SE14 where calculation-to-measurement ratio slightly exceeds four. It is seen also that calculated values of air concentrations are less than measured ones at site CZ3 located in central Europe, whereas at the rest sites calculated values exceed measured ones about two times on the average. This overestimation can be conditioned by the uncertainties of emission data (emission totals, spatial distribution of emissions and their seasonal variations). Measurements of γ-HCH concentrations in precipitation is available at EMEP sites BE14, CZ3, DE1, DE3, DE8, D9, IS91, NL91 and NO1. The comparison of measured and calculated annual means of γ-HCH concentrations in precipitation shows that calculated values occur to be lower than measured ones at almost all monitoring sites (Fig. 3.9b). For most of the monitoring sites calculated and measured γ-HCH concentrations in precipitation agree with measured ones within a factor of three. The exceptions are site DE3 (where concentrations are underestimated about 4 times) and IS91 (where the difference between measured and calculated values reaches 14 times). The comparison of measured and calculated deposition fluxes at EMEP sites FI96, SE12 and SE14 shows that the model strongly overestimates the values of deposition flux. This overestimation is even stronger than the overestimation of air concentrations at the corresponding sites. The above comparison shows that, similar to the above considered pollutants, the refinement of model description of dry and wet deposition is desirable for further refinement of the description of the contamination of the EMEP region by γ-HCH. The consideration of seasonal variations of air concentrations will be performed at the following measurement sites: CZ3, DE1, DE9, FI96, IS91, NO1, NO42, SE12 and SE14. Monthly means of calculated and measured PCB-153 air concentrations at the above sites are presented by plots in Fig. 3.10.

64

15 10 5

5

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

Feb

0

30

SE14

Measured

25

Calculated

20 15 10 5

It can be seen that about 65% of calculated and measured monthly averages of γ-HCH air concentrations agree within a factor of three. Correlation coefficients of calculated monthly averages with measured ones are normally in the range from 0.4 to 0.95. The exceptions are sites DE9 and NO42. The differences between measured and calculated values are conditioned by the fact that the model overestimates seasonal variations of air concentrations. This overestimation may be due to the uncertainties in model description of temperature dependence of γ-HCH physical-chemical parameters (vapour pressure and the Henry coefficient). So, it can be concluded that: 1. Model calculations overestimate air concentrations of γ-HCH two times on the average. This overestimation can be conditioned by the uncertainties of emission data and of physical-chemical properties of γ-HCH including their temperature dependence. 2. Correlation coefficients of calculated monthly averages with measured ones are normally in the range from 0.4 to 0.95. The patterns of seasonal variations of calculated and measured air concentrations are similar at most of the considered sites. However, these variations are overestimated by the model. 3. Further work is needed for the refinement of model parameterization of deposition flux of γ-HCH. For the refinement of model description of γ-HCH depositions more simultaneous measurements of air concentrations, concentrations and deposition flux are required.

Dec

Nov

Oct

Sep

0

Fig. 3.10. Comparison of monthly means of calculated and measured air γ-HCH concentrations at EMEP measurement sites

65

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

10

Aug

Dec

0

May

15

Jul

5

Calculated

Jun

10

NO1

Measured 20

May

15

Mar

Jan 25

Apr

Calculated

Feb

0

Mar

20

20

Jan

SE12

-HCH air concentrations, pg/m

3

25

Calculated

25

Dec

Nov

Oct

Sep

Aug

Jul

Jun

Apr

May

0

DE9

Measured

30

Feb

3


3

2


4

Measured

35

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan

Feb

6

Nov

Dec

Nov

Oct

Sep

0 Aug

8

Oct

2 Jul

10

Sep

4

Calculated

Aug

6

Jun

14 12

Jul

8

IS91

Measured

Jun

Calculated

May

18 16

May

NO42


3

12

Apr

0

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

0

Mar

5

Mar

1 Jan

10

Apr

2

Jan

15

Mar

3

Measured

20

Jan

4

10

25

Feb

5

Calculated

35 30

Jan

Calculated

DE1

Measured

Feb

3 3

7 6


FI96

Measured

45 40

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan 9 8

Feb

Calculated

Feb

3


3


CZ3


Measured

Feb

3


50 45 40 35 30 25 20 15 10 5 0

3.2. Integrated monitoring/modelling/emission approach This section is devoted to the development of an integrated approach to the comparison of modelling results with available measurements including the analysis of modelling results itself, monitoring data and emissions. For such an analysis the knowledge of influence function, say, for monthly averages of air concentrations 1 at the locations of measurement sites taking part in the comparison is of use. By influence function we mean a set of numbers γij determined in each grid cell (i, j) such that for any emission distribution Eij the following relation takes place

c = ∑ γ ij Eij ,

(3.1)

i, j

where c is air concentration at the considered location averaged over the given period (month). The numbers γij depend on the chosen location and on period of averaging but does not depend on the emission distribution. The examples of application of influence function to the analysis of agreement between measured and modelled B[a]P air concentrations will be presented in the second subsection of this section. B[a]P was chosen as a test substance since maximum number of measurement data are available for this pollutant. The most direct method of calculation of influence function is to solve the corresponding adjoint problem. However, the complexity of the model for solving the adjoint problem is the same as the complexity of the initial model (e. g. MSCE-POP). For the analysis of calculation/measurement agreement at sufficiently large number of sites the solution of the adjoint problem should be performed for each site separately which is time-consuming and resource-consuming process. That is why the approach using the adjoint problem being undoubtedly reasonable in scientific research can hardly be applicable in operational modelling. Below a method of approximate calculation of influence function using back trajectory approach is described. Being rather simple, this method allows quick calculation of an approximation of the influence function which can be effectively used at first stage of the analysis of the agreement between measurements and calculation results.

Back trajectory approach We begin with the construction of the exact solution to the following simple two-dimensional problem

⎧ ∂c ⎪ + (v, ∇ )c + αc = E , ⎨ ∂t ⎪⎩ c t =0 = 0

(3.2)

describing the 2D-transport of the pollution assuming zero initial data. Equation (2) involves only two environmental processes: advective transport of the pollutant along the wind with wind speed v and its removal from the atmosphere with removal rate α. Here E is the emission intensity. The equation (2) can be rewritten in the form

1

Monthly averages are chosen just to be definite; the approach described below can be applied to various periods of averaging. Similar, after slight modifications, deposition fluxes can be investigated instead of air concentrations. 66

dc + αc = E , dt

(3.3)

where d/dt is the full derivative along the trajectories of the wind speed v. The solution of equation (3.3) is given by t

c = ∫ e −α (t −τ ) E (τ )dτ .

(3.4)

0

where the integral is taken along the trajectory with the end at the point where the concentration is calculated. Remark. Formula (4) does not take into account boundary conditions. In fact, boundary conditions will not be essential for pollutants with high removal rate α. However, it is not complicated to involve the terms describing boundary conditions to formula (4). Integral (3.4) can be also considered as an integral along the back trajectory with the beginning at the considered point. Let us assume now that E is gridded emissions, that is, that the values of E are constant within each grid cell (i,j) equal to Eij. Then formula (3.4) can be rewritten as

c = ∑ Eij i, j

τ ij + Δτ ij

∫e τ

−α (t −τ )

dτ ,

(3.5)

ij

where Δτij – the time that trajectory spend within grid cell (i,j) and the sum is taken over all grid cells through which the trajectory travels (see Fig. 3.11).

Fig. 3.11. Integrating gridded emissions along the trajectory

To obtain the expression for monthly averages of air concentrations at the given point, right-hand part of formula (3.5) should be averaged over all trajectories arriving at the considered point within the month:

c=

( )

1 N c t (k ) , ∑ N k =1

(3.6)

where the sum is taken over all trajectories arriving at the considered point during the month with frequency 6 hours (according to the temporal resolution of the meteorological data used in the model). Finally, we have:

1 c = ∑ Eij ∑ N k i, j

τ ijk + Δτ ij( k )

∫e

τ ij( k )

(

−α t ( k ) −τ

)dτ = E γ . ∑ ij ij

(3.7)

i, j

The latter formula gives the expression for calculation of the coefficients γij of the influence function (see formula (3.1) above).

67

To obtain the expression for calculating the influence function for 3D problem, the following heuristic method is applied. First the coefficients of the influence functions for each model layer are calculated with their own wind speeds and removal rates. Then these coefficients are averaged with weights given by the vertical profile of the air concentrations for the considered chemical. This profile is assumed to be equal to the average vertical profile of concentrations obtained during the calculation run of MSCE-POP model. To test the above described approach, it was applied to the calculations of monthly averages of air concentrations obtained by MSCE-POP model run for B[a]P in 2008. Figure 3.12 shows the comparison of monthly averages of air concentrations calculated by the model with those evaluated with the help of the influence functions for a number of EMEP monitoring sites.

0.1 0.05

B[a]P air conc., ng/m3


Estimated

SE14

Calculated

0.3 0.25 0.2 0.15 0.1 0.05

0.6

Sep

Oct Oct

Sep

Nov

Estimated

SI8

Calculated

0.5 0.4 0.3 0.2 0.1

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

Feb

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

0 Feb

0 Jan

Dec

Nov

Oct

Sep

Jul

Jun

0.7

0.35

Dec

Oct

Sep

Jul

Aug

Jun

Apr

May May

Aug

Aug

Jan

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

0.4 B[a]P air conc., ng/m3

Apr

0.1 0.05 0

Feb

Nov

Dec

Oct

Sep

Aug

Jul

Jun

Apr

May

Mar

Feb

Dec

0.1

0.2 0.15

0 Jan

Dec

0.2

0

Nov

0.3

Calculated

Jul

0.005

0.4

Estimated

SE12

0.3 0.25

Jun

0.01

Calculated

0.5

May

0.015


Calculated

0.35 Estimated

PL5

0.6


0.02

Nov

0.7 Estimated

NO42

Mar

Jan

Jan

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

Feb

Nov

Dec

Oct

Sep

Jul

Aug

Jun

Apr

May

Mar

Feb

Jan

0

Aug

0.1

Jul

0.2

Calculated

Jun

0.3

Estimated

GB14

May

0.4

0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0

Apr

Calculated

0.5

0.025

0.1

Dec

Nov

Oct

Sep

Aug

Estimated

LV16

0.6

0

0.2

Apr

0.05

0.3

Mar

0.1

Calculated

0.4

Mar

0.7


Calculated

0.15

Mar

Jan

Nov

Dec

Oct

Sep

Jul

Aug

Jun

Apr

May

Mar

0.8 Estimated

FI96


0.2

Estimated

EE9

0.5

0

Jan

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Jan

Feb

0

Jul

0.1

Jun

0.2

Calculated

May

0.3

Estimated

DE9

Apr

0.4

0.6

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

Mar

Calculated

Feb

Jan B[a]P air conc., ng/m3

Estimated

DE8

0.5

0

Feb

Nov

Dec

Oct

Sep

Jul

Aug

Jun

Apr

May

Mar

Feb

Jan B[a]P air conc., ng/m3

0.1 0.05

0

0.6


0.2 0.15

Feb

0.1

0.2 0.15

Calculated

0.25

Feb

0.2

0.25

Estimated

DE3

0.3

Feb

0.3

Calculated

0.3

0.35

Feb

Calculated

Estimated

DE1


0.4

0.35

0


0.4

0.4 Estimated

CZ3



0.5

Fig. 3.12. Comparison of monthly means of calculated B[a]P air concentrations with those evaluated on the basis of the influence function approach at a number of EMEP measurement sites

68

The comparison of model results with evaluation of air concentrations on the basis of the influence function shows that the latter approach is suitable for evaluation of monthly averages of air concentrations with appropriate accuracy. Now we proceed with the description of applications of the above introduced notion of influence function to the analysis of agreement between calculations and modelling results. The example of application of this approach will be given in the next subsection. To begin with, we consider one location at which the sets of calculated (cmcalc) and measured (cmmeas) values of air concentrations are given for all months m = 1, … , 12. The following quantity shall be considered as a measure of the difference between calculations and measurements: 12

(

)

m m . Q = ∑ ccalc − cmeas 2

(3.8)

m =1

Substituting the value (7) of cmcalc to relation (8), we obtain the following expression for the criterion function: 2

⎞ ⎛ m ⎟ . Q = ∑ ⎜⎜ ∑ β m Eijγ ijm − cmeas ⎟ m=1 ⎝ i , j ⎠ 12

(3.9)

Here βmEij is emissions in the grid cell (i,j) in the month m, where βm is a coefficient introduced to reflect seasonal variability of emissions. The sensitivity sij of the criterion function (3.9) to the value of emissions in each grid cell (i,j) will be evaluated as the half of the derivative of this function with respect to Eij:

sij =

12 ⎛ ⎞ m 12 m m 1 ∂Q m m ⎟γ ij = ∑ β ccalc − cmeas = ∑ β m ⎜⎜ ∑ β m Eijγ ijm − cmeas γ ijm . ⎟ 2 ∂Eij m=1 ⎝ i , j m =1 ⎠

(

)

(3.10)

The sign of the right-hand part of (10) in a given grid cell shows the direction of emission change which will diminish the criterion function Q. Namely, if sij < 0, then enlarging emissions Eij in the cell (i,j) will lead to diminishing of the criterion function. Inversely, for the cells with sij > 0 diminishing emissions Eij leads to diminishing of the criterion function. The absolute value of the sensitivity sij shows the velocity of change in the values of the criterion function. Note that most interesting results may be obtained in the case when calculated and measured values of air concentrations are close to each other on the average. In this case the domain with possible emission increase do not cover the entire calculation domain. On the basis of information on sensitivities various emission scenarios can be elaborated. Primary evaluation of these scenarios can be performed with the help of formula (7), which allows calculating monthly values of concentrations corresponding to the modified emissions. Of course, calculations with the use of formula (7) are approximate and should be confirmed by calculations made on the basis of the full model. Further, if our task is to diminish disagreements between calculations and measurements at several sites, the sum of the criterion functions for each particular site can be used as a “complex” criterion function for the entire set of sites. The sensitivity of the “complex” criterion function to the change of emissions in each grid cell can be calculated as the arithmetical mean of sensitivities calculated for each particular site. It should be taken into account that under this approach diminishing of the “complex” criterion function can be accompanied by enlarging some particular criterion functions compensated by more rapid diminishing of other ones.

69

Application of back trajectory approach to the analysis of agreement between measurements and modelling results for B[a]P The sites which meet the following two criteria were chosen for the illustration of the application of back trajectory analysis: ¾

Measurement data are available for all 12 months.

¾

Measurements and calculated values are close to each other on the average.

The list of chosen sites is: CZ3, DE1, DE9, EE9, FI96, GB14, PL5 and SE12. Spatial distribution of sensitivities of particular criterion functions to emission changes are presented in Fig. 3.13.

Fig. 3.13. Sensitivities of criterion functions for particular EMEP sites (CZ3, DE1, DE9, EE9, FI96, GB14, PL5, SE12) to emission changes

70

It can be seen that almost for all considered site enlarging emissions in the United Kingdom, North Italy and, possibly, Poland should reduce the criterion function and, hence, will refine the agreement between measurements and modelling results. To illustrate this two emission scenarios for the site CZ3 are considered:

0.25 Standard

Modified

0.2 0.15 0.1 0.05

1. Emissions are enlarged four times in all grid cell with negative sensitivity and non-zero emissions. Seasonal variation of emissions is taken as assumed in MSCE-POP model.

Dec

Oct

Nov

Sep

Jul

Aug

Jun

Apr

May

Mar

Jan

Feb

0

Fig. 14. Emission seasonal variations

2. Emissions are modified as for first scenario but more pronounced seasonal variations are assumed (see Fig. 3.14).

3

0.8 Measured

0.7


Estimated

0.6

Scenario1

0.5 0.4 0.3 0.2 0.1

Estimated

0.6

Scenario2

0.5 0.4 0.3 0.2 0.1 Dec

Nov

Oct

Sep

Aug

Jul

Jun

Apr

May

Mar

b

Jan

Nov

Dec

Oct

Aug

Sep

Jul

Jun

May

Apr

Mar

Feb

a

Measured

0.7

0 Jan

0

0.8

Feb


3

The comparison of air concentrations estimated by formula (7) for these two scenarios with standard calculations and measurements is shown by plots in Fig. 3.15.

Fig. 3.15. Comparison of two emission scenarios at CZ3 with measurements and standard calculations (a) – scenario 1, (b) – scenario 2

It is seen that the application of Scenario 1 leads to the refinement of the agreement between measurement and modeling results, at least in winter months. At the same time, the refinement of seasonal variations is not sufficient. On the opposite, the application of Scenario 2 (which include the modification of emission seasonal variations) refines the agreement between calculation results and measurement data essentially. This once more confirms that the modification of seasonal variations can improve model performance for the considered pollutant. We stress that this section contains just an illustration of the possibilities of usage back trajectory analysis for the investigation of agreement between measurements and calculation results. More detailed analysis of the agreement for several contaminants will be performed in future.

71

3.2. Sensitivity of POP model to application of particle size-segregated deposition parameterization The current version of the MSCE-POP regional model treats particulate phase of POPs as monodisperse fraction with fixed diameter [Gusev et al., 2005]. The main purposes of the research described in this section is to evaluate the feasibility of operational modelling of POPs atmospheric transport taking into account particle size distribution and to estimate the sensitivity of calculation results to the diameter of carrier particles. This research is a continuation of work on the sizesegregated approach for modelling of POPs dispersion and removal that started at the MSC-E a year ago [Gusev et al., 2009]. To examine the influence of particle size distribution on deposition processes experimental calculations of the atmospheric transport of benzo[a]pyrene within the EMEP domain have been carried out for 2008. B[a]P has been chosen for numerical investigations because of its considerable particle-bound fraction. W For test calculations wet and dry deposition 1.0x10 schemes have been changed to size-segregated ones. The parameterization of aerosol dry deposition of MSCE-HM regional model [Travnikov and Ilyin, 2005] has been used. For wet deposition of B[a]P a 1.0x10 simple expression of the dependence between washout ratio in the lowest model layer and aerosol particle diameter has been employed. It is quasi1.0x10 linear function in double logarithmic scale (Fig. 0.1 1 10 3.16). This expression is based on literature data on dp, μm aerosol wet deposition parameterizations [Tsyro and Fig. 3.16. Washout ratio (W) as a function of Erdman, 2000; Sportisse, 2007, etc.]. For the aerosol particle diameter (dp) particle diameter dp = 0.84 μm adopted in the current monodisperse version of MSCE-POP regional model for B[a]P the new parameterization gives the same value of washout ratio (W= 5*104) as the old one. Minimum at dp = 0.2 μm (see Fig. 3.16) imitates so1.4 called “Greenfield gap” – a well known feature of 1.2 below-cloud washout consisting in weaker 1.0 scavenging of the particles with intermediate 0.8 diameters (dp ~ 0.05 - 1 μm). 0.6 6

5

dM/d(log(dp))/M

4

0.4

0.2 Several runs of the MSCE-POP model with different 0.0 parameterizations of deposition (old and new) and 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 d ,μm different B[a]P particle phase size distributions (mono-disperse and size-segregated) have been Fig. 3.17. Initial B[a]P mass distribution on done (Table 3.3). The variants with fixed particles particle sizes(dp) adopted in this work diameter dp = 0.84 μm (## 1,2 in Table 3.3) have been considered to assess the effect of deposition parameterization change just as it is. For the “size-segregated” run (# 3) it was supposed that at the moment of emission number concentration of carrier particles have lognormal distribution with mass median diameter MMD = 0.84 μm and standard deviation σ = 2. The continuous lognormal function was approximated by a quasi-constant 5-bin distribution as it shown in Fig. 3.17. The following simplified assumptions have been done for this model run: p

72

1. The boundary particle diameters for each bin are fixed (Table 3.4). 2. B[a]P mass is distributed uniformly within each bin at each time, so MMDs are constant (Table 3.4). 3. The velocities of removal processes for all particles in each bin are determined on the base of its MMD. 4. There is no mass exchange between the bins. The parameterization of the processes determined the growth of aerosol particles (coagulation, condensation, etc.) is planned to be included in the MSCE-POP model in the future. To investigate the sensitivity of calculation results to the diameter of carrier particles five scenarios with mono-disperse aerosol distribution have been examined (## 4-8). For these runs particles diameters dp have been chosen equal to MMDs of five bins mentioned above (Table 3.4). Characteristics of model runs

Table 3.3.

Name of run

Deposition parameterization

1

Base

old

2

New removal schemes

new

Mono-disperse with dp = 0.84 μm

3

Size-segregated

new

Quasi-constant 5-bin (Fig. 3.17) with MMD = 0.84 μm

4-8

New removal schemes

new

Mono-disperse with dp = 0.10 μm; 0.32 μm; 0.71 μm; 1.58 μm; 5.00 μm

N

Particle size distribution Mono-disperse with dp = 0.84 μm

Table 3.4. Characteristics of aerosol bins Number of bin 1 2 3 4 5

Boundaries, μm 0.05 – 0.20 0.20 – 0.50 0.50 – 1.00 1.00 – 2.50 2.50 - 10.00

MMD, μm 0.10 0.32 0.71 1.58 5.00

The spatial distributions of annual mean concentrations in the ambient air (a) and total annual deposition of particulate phase (b) of B[a]P in 2008 for the base run are given in Fig. 3.18. They seem traditionally and coincide with the same distributions for previous years presented and discussed in earlier MSCE reports [e.g. Gusev et al., 2009].

a

b

Fig. 3.18. Annual mean concentrations in the ambient air (a) and total annual deposition of particulate phase (b) of B[a]P in 2008 for the base model run

73

The change of parameterization of deposition (model run #2) results in the increase of total deposition flux of B[a]P over the major part of the EMEP domain (Fig. 3.19a) as a consequence of the growth of dry deposition velocities to the most types of land-cover (wet deposition velocities are equal to each other in the runs ## 1, 2 because washout ratio has one and the same value for dp = 0.84 μm – see the description of new wet deposition parameterization presented above). However, dry deposition velocity over urban areas decreases if new parameterization is used. Therefore, in regions with high fraction of urban area (e.g., the United Kingdom, east of Ukraine, Belgium - see Fig. 3.20) total deposition flux decreased. The inclusion of particle size distribution (model run #3) leads to slight increase of total deposition to the most part of the domain (compare Fig. 3.19a and b). Air concentrations calculated in the experimental model runs (## 2, 3) decreased compared to the results of the base run over the most part of the domain (Fig. 3.21). As a whole the spatial distributions of air concentrations and deposition fluxes of B[a]P for the size segregated run (#3) do not differ from those for mono-disperse run with dp = 0.84 μm (#2) considerably. As a rule, the differences between calculated values of air concentrations and depositions obtained in the base run and analogous results of size-segregated run do not exceed 20% for the most polluted regions of the Europe.

a

b

Fig. 3.19. Differences in spatial distributions of annual total deposition flux of particle B[a]P over the EMEP domain in 2008: a – difference between the run with new parameterization of deposition (#2) and the base run (#1); b – difference between the size segregated run (#3) and the base run (#1)

The comparison of the calculated results with the EMEP measurements of B[a]P concentrations in air and precipitation has not shown clear advantage of the new size-segregated deposition schemes compared to the currently used ones (Fig. 3.20, Table 3.5 – lines 1-3). It can be mentioned that the new parameterization of dry deposition gives better spatial correlation of annual mean modelled and measured B[a]P concentrations in air than the old one (Rcorr = 0.52 vs. 0.48 for the base case). Fig. 3.20. Spatial distribution of grid

74

a

b

Fig. 3.21. Differences in spatial distribution of annual air concentration of B[a]P over the EMEP domain in 2008: a – difference between the run with new parameterization of deposition (#2) and the base run (#1); b – difference between the size segregated run (#3) and the base run (#1) 0.6 Air concentration, ng/m

3

Observed

0.5

Modelled - base run Modelled - size-segregated run

0.4 0.3 0.2 0.1

b

Observed

15

2

Modelled - base run Modelled - size-segregated run

10 5 0

FI96

Observed Modelled - base run Modelled - size-segregated run

16

20

NO42

NO1

DE3

SE14

LV16

SE12

DE1

LV10

DE8

ES8

25 Deposition flux, ng/m /day

Concentration in prec., ng/L

a

GB14

EE9

DE9

SI8

CZ3

PL5

0

14 12 10 8 6 4 2 0

PL5

DE9

DE8

CZ3

DE3

DE1

SE12

SE14

FI96

c Fig. 3.22. Comparison of calculation results with measurements for B[a]P: (a) – air concentrations, (b) – concentrations in precipitation, (c) – deposition flux (annual averages)

Table 3.5.

Statistics for different model runs based on yearly mean values

N

Name of run

1 2 3 4 5

Base New removal schemes Size-segregated

6

MMD = 0.10 μm (Bin#1) MMD = 0.32 μm (Bin#2) MMD = 0.71 μm (Bin#3)

7 8

MMD = 1.58 μm (Bin#4) MMD = 5.00 μm (Bin#5)

Concentration in air , ng/m3 Rcorr(spatial) 0.007 0.48 -0.011 0.52 -0.014 0.51 0.030 0.49 0.019 0.50

Concentration in precipitation , ng/L Rcorr(spatial) 1.05 0.54 0.22 0.54 0.67 0.54 -0.34 0.52 -2.66 0.53

-0.005

0.51

-0.42

0.54

-0.035 -0.079

0.52 0.52

2.78 7.42

0.55 0.50

75

The dependence of modelling results on the carrier particle diameter (dp) has been investigated (model runs ## 4-8 with mono-disperse particle size distribution). The increase of aerosol diameter results in the considerable growth of dry and wet deposition of particle-bound B[a]P all over the model domain (Fig. 23, 24). Total dry deposition increases nearly five times with changing dp from 0.1μm to 5μm (Fig. 26), wet deposition – nearly 3 times. The minimum of wet deposition at dp= 0.32 (Fig. 26) can be explained by the presence of minimum in washout ratio curve (Fig. 3.16). At the same time, air content decreases (Fig. 25). Total mass of particle phase of B[a]P in air falls five times (Fig. 27). It should be mentioned, that the form of all the spatial distributions does not change essentially (Fig. 3.23-3.25).

a b c Fig. 3.23. Annual dry deposition of particle-bound B[a]P for three model runs with different particle MMD: a – dp =0.1 μm (bin #1), b - dp =0.7 μm (bin #3) , c - dp =5.0 μm (bin #5)

a b c Fig. 3.24. Annual wet deposition of particle-bound B[a]P for three model runs with different particle MMD: a – dp =0.1 μm (bin #1), b - dp =0.7 μm (bin #3) , c - dp =5.0 μm (bin #5)

a

b

c

Fig. 3.25. Annual mean concentrations of B[a]P in the ambient air for three model runs with different particle MMD: a – dp =0.1 μm (bin #1) , b - dp =0.7 μm (bin #3) , c - dp =5.0 μm (bin #5)

76

35 30

Normalized total air content, %

Percent of emissions

40

Wet dep. Dry dep.

25 20 15 10 5 0 0.10

0.32

0.71

1.58

100 90 80 70 60 50 40 30 20 10 0

5.00

0.10

Particle diameter, μm

0.32

0.71

1.58

5.00

Particle diameter, μm

Fig. 3.26. Annual wet and dry deposition of aerosol phase of B[a]P to EMEP domain in percent of emitted mass The statistical indicators for the bin runs (five lower lines selected by a color in Table 2) differ from each other in the following way: the increasing of dp from 0.1μm to 5μm results in the fall of yearly mean B[a]P concentrations in air averaged by the EMEP stations (BIAS changes from 0.030 ng/m3 to -0.079 ng/m3) and the growth of concentrations in precipitation (BIAS changes from -0.34 ng/L to 7.42 ng/L). Spatial correlation of concentrations in air and precipitation does not change noticeably.

dM/d(log(dp))/M

Fig. 3.27. Annual mean total aerosol phase mass of B[a]P in the whole EMEP domain normalized to the results of the model run with dp =0.01μm

1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 1.E-02

Annual mean air conc at EMEP grid sell (43,86) Emission

1.E-01

1.E+00 dp,mm

1.E+01

1.E+02

Fig. 3.28. Particle size distributions of lead mass Turning back to the size segregated run one could in the EMEP grid cell (43,86) in comparison with say that particle size distribution of B[a]P mass emission distribution changes according to the regularity ascertained above. Namely, the contribution of the mass transported on fine particles increases while the corresponding contribution of coarse particles decreases, i.e. MMD diminishes. As an example, calculated annual mean B[a]P mass distributions in the EMEP grid cell (43,86) (Norway) in comparison to the initial distribution (at the moment of emission) are given in Fig. 3.28. So, several numerical experiments on size-segregated modelling of B[a]P particle phase have been done at the MSC-E. The following main results have been obtained. 1. The parameterizations of washout ratio and dry deposition velocities of aerosols to different types of land-cover have been changed for experimental modelling of B[a]P long-range transport and deposition within the EMEP region. The new parameterizations are size-dependent in contrast to the old ones. 2. The replacement of deposition schemes have resulted in the growth of total deposition flux of particulate phase of B[a]P to the most part of the EMEP region (except of some areas with high fraction of urban type of land-cover). 3. The comparison of the modelling results with the EMEP measurements of concentrations in air, in precipitation, and wet deposition of B[a]P has not shown little improvement of the new deposition schemes with respect to the old ones. Experimental numerical modelling of size-resolved POP dispersion and removal is planned to be continued. Further improvement of the schemes of particle dry and wet deposition in the MSCE-POP model and further investigations of the size distribution of aerosol particles carrying POPs are planned to be performed in the future.

77

CONCLUSIONS Sensitivity to meteorological parameters associated with climate change 1. Pilot version of the approach aimed at the evaluation of sensitivity of HMs and POP pollution levels to meteorological and geophysical parameters has been developed and tested for B[a]P, β-endosulfan and lead. Most important parameters (factors) affecting atmospheric dispersion and deposition of the pollutants were determined. The directions of further elaboration of the approach were outlined. 2. For B[a]P most important meteorological parameters determining its long-range transport are temperature and vegetation cover. For particular locations, precipitation intensity also significantly affects long-range transport of B[a]P. In case of β-endosulfan one of the key parameters is fraction of area covered by water bodies. 3. Sensitivity of lead deposition in countries to meteorological parameters (precipitation, temperature, wind velocity, friction velocity) was analysed for the Czech Republic, Italy, the United Kingdom and Finland. Precipitation amounts and wind velocity were among the most important meteorological parameters controlling deposition levels in countries. In the framework of the further development of this approach list of meteorological factors should be specified.

Heavy metals 1. Deposition of lead and cadmium in Europe and Central Asia as a whole in 2008 was about 14% and 10% lower than that in 2007, respectively. Mercury deposition to the EMEP region remained almost the same. Decrease of lead and cadmium deposition in the southern, central and western parts of Europe was caused by the decline of emission values and resuspension. Increase of precipitation amounts caused the rise of deposition of lead and cadmium in the western part of the United Kingdom and the central part of Kazakhstan and of mercury in the north-western Italy, Russia and most of Scandinavia. Deposition of mercury in Denmark, Cyprus, France, Norway and Germany decreased, whereas in Slovakia and Romania increased because of the corresponding changes of the emissions. 2. Quality of the air pollution assessment was characterized by means of integrated approach taking into account uncertainties of the model, emission data and measurements. The uncertainty of country totals of heavy metal emission typically ranged between 30 – 60%. Overall uncertainty of measured wet deposition estimated using results of field campaigns was around 20% for lead and cadmium, and 40% of mercury. However, these estimates did not include the effect of representativeness of stations location. 3. Modelling results agreed with measurement data with satisfactory accuracy, keeping in mind uncertainties of the emission and monitoring data. At around 2/3 of stations calculated and observed concentrations of lead and cadmium in air agreed within ±50%. Modelled wet deposition fluxes matched the observed ones within ±50% at about 2/3 of stations for lead and about half of stations for cadmium.

78

4. The discrepancy between the simulated and observed mercury concentrations did not exceed ±15% for air concentrations. For concentrations in precipitation and wet deposition fluxes the bias was within ±30% for most of stations. 5. Annual set of meteorological data with fine spatial resolution was prepared for the Czech Republic (10x10 km2, 5x5 km2) and Croatia (10x10 km2) in the framework of the EMEP Case Study on heavy metal pollution assessment. The generated meteorological parameters were compared with observational data. 6. The increase of spatial resolution leads to the improvement of statistical indexes for the Czech Republic and Croatia. Spatial Root Mean Square Error decreases (improves) and correlation coefficient increases for all analysed meteorological parameters (temperature at 2 m, wind speed at 10 m, precipitation amount) for the Czech Republic and Croatia as resolution increases. The same is true for bias except for precipitation amounts over the domain for the Czech Republic. Improvement of the indexes related to temporal variability is smaller compared to spatial indexes. 7. Applicability and peculiarities of use of auxiliary measurements for the assessment of pollution levels and the model validation were investigated in cooperation with the ICP-Vegetation of WGE. Both wet and dry deposition should be considered when concentrations in mosses and deposition are compared. The agreement between the two types of data increased in regions with high density of measurements of concentrations in mosses.

Persistent Organic Pollutants 1. Calculated values of B[a]P air concentrations agree with measurements within a factor of two at about 60% of measurement sites, and within a factor of three – at about 75% of sites. At the same time calculations exceed measurements 1.7 times on the average. Temporal coefficients of correlation between monthly mean modelled and observed concentrations lie within 0.6 – 0.95 range. Amplitude of seasonal variations was somewhat underestimated by the model. 2. Annual means of PCB-153 air concentrations agree with measured ones within a factor of two. On average, the model underestimates the observed concentrations by about 20% on average. Significant coefficient of correlation (0.89) between the modelled and measured values indicates that spatial distribution of PCB-153 air concentrations is well captured by the model. Temporal correlation coefficient for monthly mean values ranges within 0.64 – 0.9 for most of stations. Magnitude of peaks of the observed concentrations is typically underestimated by the model. 3. Air concentrations of γ-HCH are overestimated by the model about twice. Correlation coefficient for monthly mean values of modelled and observed concentrations varies from 0.4 to 0.95. At most of stations amplitude of seasonal variability simulated by the model significantly exceeds that based on the measurement data. 4. For the further progress in modelling of POPs the refinement of the model deposition schemes is needed. Information on seasonal variability of PCB-153 emissions is required. More colocated measurements of concentrations in air, in precipitation and deposition fluxes of γ-HCH could also significantly contribute to the improvement of the POP modelling approaches.

79

5. The particle size-depended parameterizations of washout ratio and dry deposition velocities of aerosols were developed and tested for B[a]P. Size-segregated deposition schemes resulted to the growth of total deposition flux of particulate phase of B[a]P to the most part of the EMEP region. The comparison of the modelling results with the EMEP measurements of concentrations in air, in precipitation, and wet deposition of B[a]P exhibit little improvement compared to the results based on monodisperse consideration of particles.

80

REFERENCES Aas W. [2006]. Data quality 2004, quality assurance and field comparisons. EMEP/CCC Report 4/2006. Aas W., Alleman L.Y., Bieber E., Gladtke D., Houdret J.L., Karlsson V., Monies C. [2009]. Comparison of methods for measuring atmospheric deposition of arsenic, cadmium, nickel and lead. J. Environ. Monit., vol. 11, pp. 1276 - 1283 doi: 10.1039/B822330K. CITEPA [2010] Inventaire des emissions de polluants atmospheriques en France au titre de la convention sur la pollution atmospherique transfrontaliere a longue distance et de la directive Europeenne relative aux plafonds d’emissions nationaux (NEC). CEE–NU/NFR & NEC Mars 2010. p.221. Dudhia J. [1996] A multi-layer soil temperature model for MM5. Preprints, The Sixth PSU/NCAR Mesoscale Model Users' Workshop, 22-24 July 1996, Boulder, Colorado, pp.49-50. Grell G., Dudhia J. and Stauffer D. [1994] A Description of the Fifth Generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note, TN-398+STR, p.117. Gusev A., Mantseva E., Shatalov V., and B.Strukov [2005] Regional multicompartment model MSCE-POP. MSC-E Technical Report 5/2005. Gusev A., O. Rozovskaya, V. Shatalov, V.Sokovykh [2009] Persistent Organic Pollutants in the Environment. EMEP Status Report 3/2009. Harmens H. and Norris D. and participants of the moss survey [2008]. Spatial and temporal trends in heavy metal accumulation in mosses in Europe (1990 - 2005). Report of ICP Vegetation, 51 pp. Hong S.-Y. and Pan H.-L. [1996] Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124, pp.2322-2339. Ilyin I., O. Rozovskaya, V. Sokovyh, O. Travnikov, M. Varygina, W. Aas, H.T. Uggerud [2010a] Heavy metals: Transboundary Pollution of the Environment. EMEP Status Report 2/2010. Kain J. S. [2002] The Kain-Fritsch convective parameterization: An update. J. Appl. Meteor., 43, pp.170-181. Kain J. S. and Fritsch J. M. [1993] Convective parameterization for mesoscale models: The Kain-Fritsch scheme. The representation of cumulus convection in numerical models, K. A. Emanuel and D. J. Raymond, Eds., Amer. Meteor. Soc., p.246. Murrells T.P., Passant N.R., Thistlethwaite G., Wagner A., Li Y., Bush T., Norris J., Whiting R., Walker C., Stewart R.A., Tsagatakis I., Conolly C., Brophy N.C.J. and S. Okamura [2010] UK Informative Inventory Report (1970 to 2008). UK Emissions Inventory Team, AEA Group, p.50. Nielsen O.-K., Winther M., Mikkelsen M.H., Hoffmann L., Nielsen M., Gyldenkaerne S., Fauser P., Plejdrup M.S., Albrektsen R. and K.Hjelgaard [2010] Annual Danish Informative Inventory Report to UNECE. Emission inventories from the base year of the protocols to year 2008. NERI Technical Report no.774, p.34. Reisner J., Rasmussen R. J., and Bruintjes R. T. [1998] Explicit forecasting of supercooled liquid water in winter stroms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124B, pp.1071-1107. SEPA [2010] Informative Inventory Report 2010. Sweden. Annexes. Swedish Environmental Protection Agency, p.74. Sportisse B. [2007] A review of parameterizations for modelling dry deposition and scavenging of radionuclides. Atmos. Environ. vol.41, pp.2683-2698. SYKE [2010] Air Pollutant Emissions in Finland 1980-2008. Informative Inventory Report. Finnish Environment Institute. Consumption and Production Centre, Environmental Performance Division Air Emissions Team, p.55. Travnikov O. and I.Ilyin [2005] Regional Model MSCE-HM of Heavy Metal Transboundary Air Pollution in Europe. EMEP/MSC-E Technical Report 6/2005, 59 pp. Tsyro S. and L. Erdman [2000] Parameterization of aerosol deposition processes in EMEP MSC-E and MSC-W transport models. EMEP/MSC-W Technical Note.

81