ANNEX I

DETAILED SPECIFICATION OF THE PRESENTED AIR POLLUTION MAPS

The spatial maps are based on the results of measurements taken at individual localities and are created by combining information from various sources (Chapter XII). The uncertainties of individual maps depend particularly on the density of measuring stations and the equivalent coverage across the Czech Republic, on the uncertainties of individual measurements, inputs into models, model calculations and the method used for the construction of the spatial maps. The greatest uncertainty of the maps is in the vicinity of the measuring stations. In spite of the rather high uncertainties of some of the maps, the air pollution field estimates appropriately correspond to the information used and the current state of the art. When interpreting the maps, it is necessary to take into account their uncertainties.

The following paragraphs present the background material used in the construction of air pollution maps for the year 2014 and the specification of individual maps presented in this yearbook.

1. Data used

1. Measured air pollution data: Annual characteristics of the measured data from the ISKO database are used.
2. Outputs of dispersion models: Outputs of the following models are used:
   SYMOS – Gaussian model, resolution 2×2 km, year 2014 (meteorology: wind roses 2014, emissions: REZZO 2013);
   CAMx – Eulerian model, resolution 4.7×4.7 km, year 2014 (meteorology: ALADIN 2014, emissions: REZZO 2010 for the area of the Czech Republic, TNO 2009 for the surrounding area); and
   EMEP – Eulerian model, resolution 50×50 km, year 2012 (meteorology: ECWMF 2012, emissions: EMEP 2012).
   The individual models used the latest outputs available during the preparation of this yearbook.
3. Emissions from traffic: resolution 1x1 km, source: emission database REZZO 4 (year 2013).
4. Altitude: resolution 1x1 km, source: ZABAGED, Office for Surveying and Mapping.
5. Population density: resolution 1×1 km, source: CSO.
    

2. The estimate of uncertainty

The uncertainty of each respective map was estimated using the cross-validation method, see Horálek et al. (2007). Estimates of concentrations for each measuring site are always produced by omitting the given measurement and using other data. They thus objectively represent the quality of the map beyond the measuring site. This procedure was used repeatedly for all measuring sites. The estimated values were compared with the measured values with the use of the root-mean-square error (RMSE) or the relative root-mean-square error (RRMSE):
 

                  ,

where

        Z(s_i)  is the measured value of the concentration in the i-th point,
          is the estimate in the i-th point using other data,
        N       is the number of measuring stations.

Due to computation reasons, the estimate of uncertainty was calculated only for the interpolation of residuals. The total uncertainty of the map is therefore slightly higher. It is also necessary to mention the fact that this is the mean uncertainty of the whole map. The spatial distribution of uncertainty was not estimated.

3. Parameters of individual maps

The following tables present the supplementary quantities used in the linear regression model and their parameters (c, a1, a2, …), parameters of interpolation with the use of kriging (range, nugget, partial sill) and inverse distance weighting (IDW weight). In most maps also the RMSE is presented. The parameters are presented for individual layers (rural, urban, traffic).

1. Suspended particles PM₁₀:
   The maps were created using measurements from 37 rural, 100 urban and suburban background stations, and 32 traffic stations. The results of twelve industrial stations were considered in their immediate vicinity only (Table 1).
    
2. Fine suspended particles PM_2.5:
   The map was created using measurements from 15 rural and 29 urban and suburban background stations. The results from nine traffic and four industrial stations were considered in their immediate vicinity only. Due to the mapping methodology, the uncertainty of the map was not calculated (Table 2). The PM₁₀ map is used as a supplementary quantity – considering the strong regression relationship between PM₁₀ and PM_2.5, the uncertainty estimate would be underestimated.
    
3. Benzo[a]pyrene:
   The map was created using measurements from 3 rural and 27 urban and suburban background and traffic stations. The results from four industrial stations were considered in their immediate vicinity only. Because of the very low number of rural stations, the estimate of uncertainty in rural areas is only approximate. The low number of rural stations is also the cause of the relatively high uncertainty of the map in rural areas (Table 3).
    
4. Nitrogen dioxide and nitrogen oxides:
   Nitrogen dioxide and nitrogen oxides: The maps were created using measurements from 17 rural stations, 41 urban and suburban background stations, and 25 traffic stations. The results from 17 industrial stations were considered in their immediate vicinity only (Table 4).
    
5. Ground-level ozone:
   The maps were created on the basis of measurements from 32 rural stations, 45 urban and suburban background, and 6 traffic stations (Table 5).
    
6. Benzene:
   The map was created using measurements from 23 background stations (4 rural, 19 urban and suburban) and 5 traffic stations. The results from 3 industrial stations were considered in their immediate vicinity only (Table 6).
    
7. Heavy metals:
   The maps were created using measurements from 10 rural and 43 urban and suburban stations (without distinguishing background, traffic and industrial ones). The uncertainty of the cadmium map was estimated without data from the Tanvald station and its immediate vicinity because high absolute values of concentrations at this locality would distort the total uncertainty of the map. The high relative uncertainty of the cadmium map is connected with low values of cadmium in most of the country (Table 7).
    
8. Sulphur dioxide:
   The maps were created using measurements from 26 rural stations (without distinguishing background and industrial ones) and 31 urban and suburban background stations. The results from 4 traffic and 6 industrial stations were considered in their immediate vicinity only (Table 8).
    

The numbers of stations also comprise foreign (German and Polish) stations whose measurements were used to construct the maps.

For the merging of rural and urban layers, the margins of classification intervals (see Chapter XII.) were set to: a₁ = 200 inhbs.km^-2, a₂ = 1000 inhbs.km^-2. For the merging of background and traffic layers, the margins of classification intervals were set to: a₁ = 1.5 t[PM₁₀].year^-1.km^-2, a₂ = 5 t[PM₁₀].year^-1.km^-2 (for PM₁₀), resp. a₁ = a₂ = 10 t[NO₂].year^-1.km^-2 (for NO₂, NO_x and O₃).

Tab. 1 Parameters of PM₁₀ maps

Tab. 2 Parameters of PM_2.5 map

Tab. 3 Parameters of benzo[a]pyrene map

Tab. 4 Parameters of NO₂ and NO_x maps

Tab. 5 Parameters of ground-level ozone maps

Tab. 6 Parameters of benzene map

Tab. 7 Parameters of arsenic and cadmium maps

Tab. 8 Parameters of SO₂ maps