# Input data for models of MSCE Ilia Ilyin

- Slides: 26

Input data for models of MSC-E Ilia Ilyin EMEP/MSC-E 6 th TFMM Meeting, Zagreb, April 2005 1 EMEP/MSC-E

MSCE-HM&POP model Pollutant properties Geophysical data Emissions Meteorological data MSCEHM/POP Model output EMEP/MSC-E 6 th 2 TFMM Meeting, Zagreb, April 2005

Geophysical data - Land-use (18 categories) - Leaf Area Index - Organic content of soil - Chemical reactants - Topography - etc …. EMEP/MSC-E 6 th 3 TFMM Meeting, Zagreb, April 2005

Emission data Within EMEP 44 countries Available official data: - National totals: 11 - 35 countries - Spatial distribution: 18 countries (HMs), 10 (POPs) - Emission sector data: 23 countries - Seasonal variability: no official data - Uncertainty analysis: Denmark only Official data are incomplete Expert estimates are used extensively Natural emission, re-emission of HMs: the data are prepared by MSC-E EMEP/MSC-E 6 th 4 TFMM Meeting, Zagreb, April 2005

Meteorological data üMeteorological preprocessor üMeteorological variability üReference meteorology EMEP/MSC-E 6 th 5 TFMM Meeting, Zagreb, April 2005

Meteorological data - For hemispheric model: SDA (System of Diagnosis of low Atmosphere) , Hydrometeorological Centre of Russia - For regional version of the model: MM 5 system (Pennsylvania State University/NCAR Community model) EMEP/MSC-E 6 th 6 TFMM Meeting, Zagreb, April 2005

Meteorological pre-processor: MM 5 (version 3) MM 5 features: • Support various map projections, including stereographic one • Support various data sets, e. g. re-analysis of PSU/NCAR mesoscale model NCEP/DOE or ECMWF • Parameterizations selected by user • Works reasonably fast • Available nesting • World-wide spread and tested EMEP/MSC-E 6 th 7 TFMM Meeting, Zagreb, April 2005

MM 5 adaptation for EMEP tasks § Vertical structure: Like in transport model § Horizontal structure: EMEP grid + surrounding 6 gridcells § 3 D precipitations are introduced by MSC-E § NCEP/DOE Reanalysis data EMEP domain EMEP/MSC-E MM 5 domain 6 th 8 TFMM Meeting, Zagreb, April 2005

Parameters derived form MM 5 (1990 - 2002) Temporal resolution: 6 hours - Horizontal wind components Surface pressure Air temperature Water vapour mixing ratio Liquid water mixing ratio Ice mixing ratio Convective precipitation Large-scale precipitation Turbulent coefficient Surface temperature Monin-Obukhov length scale Friction velocity Snow cover height EMEP/MSC-E 6 th 9 TFMM Meeting, Zagreb, April 2005

Specific tasks (effect of meteo variability is important): - Long-term pollution trends in response to emission reduction - Future emission projections - Critical loads approach and risk assessment EMEP/MSC-E 6 th 10 TFMM Meeting, Zagreb, April 2005

Effects of meteorological variability on modelling results - Pb and Hg - 1990 – 2002 - Constant emission - Concentrations in air, in precip, total depositions Relative deviation: Yi, j - model output parameter EMEP/MSC-E 6 th 11 TFMM Meeting, Zagreb, April 2005

Relative deviations (air concentrations, Pb) Lead Map of relative deviation EMEP/MSC-E 6 th 12 TFMM Meeting, Zagreb, April 2005

Model output uncertainty caused by meteorological variability (Pb) MSRE - Mean-Square Relative Error bars: range (90% interval) MSRE = 10 - 45% for Pb 6 th 13 TFMM Meeting, Zagreb, April 2005

Relative deviations (TGM air concentrations) TGM Map of relative deviation EMEP/MSC-E 6 th 14 TFMM Meeting, Zagreb, April 2005

Model output uncertainty caused by meteorological variability (Hg) Error bars: range (90% interval) MSRE = 3 - 10% (TGM) = 10 - 40% (Conc. in precip, total depositions) 6 th 15 TFMM Meeting, Zagreb, April 2005

How can we minimize the effects of meteorological variability? 2 possible solutions: a) Long-term model runs with constant annual emission and further averaging b) Meaningful selection of “reference” year Reference year: A year, which computed model output parameters most resemble those averaged over long period of time EMEP/MSC-E 6 th 16 TFMM Meeting, Zagreb, April 2005

Selection of the reference year Parameters analyzed (1990 - 2002): - Air concentrations - Concentrations in precipitation - Total depositions - Precipitation annual sums Key parameters (WGE request): - Lead total depositions - Mercury concentrations in precipitation EMEP/MSC-E 6 th 17 TFMM Meeting, Zagreb, April 2005

Statistical criteria of the selection: Normalized Mean Square Error Di, j - annual depositions in (i, j) point Dmean, i, j - multi-annual mean deposition in (i, j) point - spatial mean Dmean - spatial mean D Fractional Bias Fractional Standard Deviation Correlation Coefficient Liner regression coefficients EMEP/MSC-E 6 th 18 TFMM Meeting, Zagreb, April 2005

Normalized Mean-Square Error, (Pb, total depositions) The lowest NMSE in 1990 EMEP/MSC-E 6 th 19 TFMM Meeting, Zagreb, April 2005

Difference between mean deposition field of Pb and deposition in reference year (1990) Relative deviation = (D 1990 - Dmean)/Dmean x 100% Relative deviation: ± 20% over 90% of area EMEP/MSC-E 6 th 20 TFMM Meeting, Zagreb, April 2005

Normalized Mean-Square Error, (Hg, concentrations in precipitation) EMEP/MSC-E 6 th 21 TFMM Meeting, Zagreb, April 2005

Conclusions Emission • Official data on HMs and POPs emissions are incomplete. To fill the gaps in the data, expert estimates are used Meteorology • Variability of concentrations and depositions of aerosol species ranges from 10 to 45 % due to meteorological variability • Variability of TGM concentrations ranges from 3 to 10%, and of concentrations in precipitation and depositions of Hg from 10 to 40% due to meteorological variability. • 1990 can be chosen as reference year basing on data for 1990 – 2002. Relative deviation of depositions in 1990 from multi-annual mean depositions <20% EMEP/MSC-E 6 th 22 TFMM Meeting, Zagreb, April 2005

23

Criteria to select the reference year Normalized Mean Square Error Pi, j - depositions in point (i, j) in a separate year Oi, j - depositions in point (i, j) averaged over 1990– 2002 - spatially averaged P - spatially averaged O N - number of grid cells 24

Criteria to select the reference year(2) Fractional Bias Fractional Standard Deviation σO, σP - standard deviations Correlation Coefficient Liner regression coefficients 25

Natural emission and re-emission of Pb and Cd Land Sea Pb, g/km 2/y 220 160 Cd, g/km 2/y 12 8 Snow/ice surfaces – zero emission 26