A Methodology for Seasonal Photochemical Model Simulation Assessment Veronica Gabusi and Marialuisa Volta {gabusi,lvolta}@ing.unibs.it Dipartimento di Elettronica per l Automazione Università degli Studi di Brescia - Italy
Outline Seasonal photochemical model assessment Critical issues The proposed methodology Application cases Northern Italy domains GAMES Modelling System 1996 Summer Season The CityDelta Modelling Exercise Conclusions
Seasonal Simulation Assessment Critical issues: no standard performance procedure exists each run of the models results in a large amount of output data Comparison between punctual and average values to focus the analysis for a restricted number of typical concentration patterns to summarize the simulation results over long time period by means of proper performance indexes.
The Model Evaluation Methodology (1) Clustering process (2) Sites Identification Used for classifying patterns Ozone concentrations Daily shape, peak distribution Each cluster can be represented by a single station Representative station Virtual station (3) Performance indicators Statistical indicators (EPA, EC) Other Statistical indexes O3 exposure values Percentile values Model inter-comparison
Seasonal Application Cases Simulation performed with GAMES system over Northern Italy domains Preliminary Application: 1996 Summer Season - Contribution to EUROTRAC2/SATURN Project CityDelta Modelling Exercise
ECMWF model Wind and Temperature profile EMEP model concentrations IC_BC The GAMES system Gas Aerosol Modelling Evaluation System Topography Land Use Initial and boundary conditions CALMET CALGRID 3-dimensional pollutant Concentration patterns Local measurements -Synop - Regional network 3D wind and temperature fields turbulence parameters Emission fields Upper air sounding data temperature profiles Emission model CORINAIR Emission inventory Spatial disaggregating coefficients Emission time patterns VOC splitting profiles
UTM (km) 5250 5200 5150 5100 5050 5000 4950 4900 4850 4800 250 300 350 400 450 500 550 600 650 700 750 800 UTM (km) 240km x 232km complex terrain Case I: the selected domain industrialised and populated area close road network critical anthropic emissions the simulated period: April-September 1996 (UTM -km) 5160 5140 5120 5100 5080 5060 5040 5020 5000 4980 4960 VA CO PV MI LC LO BG 460 480 500 520 540 560 580 600 620 640 660 680 (UTM - km) SO CR BS MN VR
5150.00 5100.00 5050.00 5000.00 ALRENON MOTTARONE POST_121 POST_308 POST_903 POST_902 POST_674 POST_675 POST_717 POST_602 POST_606 POST_673 POST_616 POST_624 POST_676 POST_618 POST_715 POST_698 POST_716 POST_662 POST_905 POST_907 POST_906 POST_503 POST_209 POST_801 POST_201 POST_526 POST_304 POST_301 POST_305 SPIETROCA POST_654 POST_751 POST_711 mottar 903 902 Clustering process and sites identification 0 10 20 30 40 50 Distances 301 308 304 305 674 906 907 905 673 662 675 698715 618676 616 716 602 717 606 624 654 711 751 121 801 201 209 526 503 renon Mottaron Varenna LegnLimbiate Vimercat Mi_Juva mean day [ppb] 90 80 70 60 50 40 30 20 10 Gambara Renon Rural Urban Remote Case I 4950.00 spcapof 500.00 550.00 600.00 650.00 700.00 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 GAMBARA LEGN_SMA LIMBIATE MI_JUVA MOTTARON RENON VARENNA VIMERCAT [hour]
US EPA guidelines: Statistical Indexes Mean Normalized Bias Error (MNBE), ±5 15% Mean Normalized Gross Error (MNGE), ±30 35% EC Directive: the 1 hour averages (daytime), ± 50% the 8 hours daily maximum, ± 50% Model intercomparison: CALGRID STEM 1 n n 1 = 1 n n 1 = 1 Legnano Limbiate Juvara Vimercate Ispra Parma Varenna 1 C C mod mod ( x, t) C C ( x, t) obs ( x, t) C C obs obs ( x, t) Case I obs ( x, t) ( x, t) O3-1-h daily max O3-8-h daily max NO2 - mean conc. mnbe [%] mnge [%] mnbe [%] mnge [%] mnbe [%] mnge [%] 4.4 27.6 14 32.6-29.8 29.4-15.4 27.5-11.2 26.2 2.5 32.0 1.4 31.2 14.3 36.6-20.1 25.9-5.6 28.3 1.8 28.5-17.5 33.3 13.3 25.6 19.7 29.6 - - 21.6 30.4 25.5 32.2 - - -1.3 25.2 0.6 23.1 - -
Computed 180 160 140 120 100 Computed 80 60 40 20 180 160 140 120 100 80 60 40 20 LEDA - d_1hmax 0 0 20 40 60 80 100 120 140 160 180 STEM CALGRID LEDA - d_8hmax 0 0 20 40 60 80 100 120 140 160 180 STEM CALGRID Measured [ppb] Model inter-comparison Measured [ppb] Computed Computed 180 160 140 120 100 80 60 40 20 HEDA - d_1hmax 0 0 20 40 60 80 100 120 140 160 180 180 160 140 120 100 80 60 40 20 STEM CALGRID HEDA - d_8hmax Measured [ppb] 0 0 20 40 60 80 100 120 140 160 180 STEM CALGRID Measured [ppb] Case I LEDA: low emission density area HEDA: high emission density area
Case II: the selected domain 300km x 300km the simulated period: April-September 1999
COSSATO VA_VIDOL CHIAVENN RE_MASSE PARMA PIACENZA CREMA AGRATE VIMERCAT LIMITO MEDA ARCONATE MOTTA_V MAGENTA CASTELLA Ozone pattern classification (I) Case II Cluster Tree UTM (km) 5150 5100 5050 5000 4950 4900 0 10 20 30 Distances TORINO NOVARA VARESE SONDRIO TRENTO BERGAMO BRESCIA MILANO VERONA PIACENZA Cluster 1 ALESSANDRIA Cluster 2 PARMA Cluster 3 Cluster 4 Cluster 5 GENOVA Cluster 6 400 450 500 550 600 650 UTM (km) MODENA
O3 [ppb] 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 Ozone pattern classification (II) 0 4 8 12 16 20 24 1 2 3 4 5 6 [hour] UTM (km) 5150 5100 5050 5000 4950 4900 TORINO SONDRIO VARESE BERGAMO BRESCIA NOVARA MILANO VERONA PIACENZA Cluster 1 ALESSANDRIA Cluster 2 PARMA Cluster 3 Cluster 4 Cluster 5 GENOVA Cluster 6 400 450 500 550 600 650 UTM (km) Case II TRENTO MODENA
O3 [ppb] 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 Representative station definition -Representative station new clustering process -Virtual station averaging, hour by hour, the cluster 1 cluster 3 cluster 5 Case II 0 4 8 12 16 20 24 0 4 8 12 16 20 240 4 8 12 16 20 24 Varese_obs Varese_comp average_obs average_comp [hour] measurements recorded in the stations belonging to the group Vimercate_obs average_obs [hour] Magenta_obs average_obs [hour] Vimercate_comp average_comp Magenta_comp average_comp Rural cluster Urban cluster Suburban cluster
Performance evaluation Model evaluation (I) Graphical analysis: time series, scatter plots, Statistical indicators: US EPA guidelines: Mean Normalized Bias Error (MNBE), ±5 15% Mean Normalized Gross Error (MNGE), 30 35% Max 1h Max 8h Cluster MNBE MNGE R RMSE MNBE MNGE R RMSE 1-7 22 0.37 17.52-2 11 0.41 14.05 2 13 26 0.6 14.48 28 13 0.61 15.34 3-10 25 0.63 20.11-3 15 0.64 16.47 4-18 23 0.62 22-14 12 0.65 17.66 5 15 33 0.59 17.24 19 16 0.69 14.15 Case II O3 concentrations
Observed 120 100 80 60 40 20 0 Model evaluation (II) 0 20 40 60 80 100 120 Sim ulated 8h p25 8h p50 8h p75 8h p95 120 100 80 60 40 20 0 Urban cluster 0 20 40 60 80 100 120 Simulated 1h p25 1h p50 1h p75 1h p95 1h p25 1h p50 1h p75 1h p95 (repres.) 1h p25 1h p50 1h p75 1h p95 (virtual) Case II
Conclusions Methodological approach to evaluate seasonal photochemical simulation Simulations performed in complex domains Main issues: 1996 Summer Season (SATURN) model inter-comparison 1999 Summer Season (CityDelta Project) Statistical and graphical methods Each station or representative station The evaluation of representative stations should be preferred with long-term simulation assessments