Comparison of air quality model results with urban ... · Budapest during the winter 2018-2019....
Transcript of Comparison of air quality model results with urban ... · Budapest during the winter 2018-2019....
Comparison of air quality model
results with urban
measurements in Budapest, Hungary
Ádám Leelőssy, Attila Kovács, István Lagzi, Adrienn Balogh, Róbert Mészáros
Department of Meteorology, Eötvös Loránd University, Budapest
Problem statement
Problem:
High and frequent PM10 air pollution in the winters in Budapest
Public and political focus on forecasting & understanding
Challenges:
Very sensitive on local sources (domestic heating)
Emissions are very uncertain and variable (domestic heating)
Large intra-urban variability of air quality
Low-level inversions
Possibilities:
Good observed data availability (12 sites within Budapest with hourly samples)
Temporal stationarity of smog events (several days of anticyclonic weather)
Are smog events predictable by CAMS models?
Is the CAMS model error smaller than intra-urban variability?
Is there a „best” CAMS model for Budapest?
Is CAMS better than naive (persistence) forecasting?
Problem statement
Problem:
High and frequent PM10 air pollution in the winters in Budapest
Public and political focus on forecasting & understanding
Challenges:
Very sensitive on local sources (domestic heating)
Emissions are very uncertain and variable (domestic heating)
Large intra-urban variability of air quality
Low-level inversions
Possibilities:
Good observed data availability (12 sites within Budapest with hourly samples)
Temporal stationarity of smog events (several days of anticyclonic weather)
Are smog events predictable by CAMS models?
Is the CAMS model error smaller than intra-urban variability?
Is there a „best” CAMS model for Budapest?
Is CAMS better than naive (persistence) forecasting?
Problem statement
Problem:
High and frequent PM10 air pollution in the winters in Budapest
Public and political focus on forecasting & understanding
Challenges:
Very sensitive on local sources (domestic heating)
Emissions are very uncertain and variable (domestic heating)
Large intra-urban variability of air quality
Low-level inversions
Possibilities:
Good observed data availability (12 sites within Budapest with hourly samples)
Temporal stationarity of smog events (several days of anticyclonic weather)
Are smog events predictable by CAMS models?
Is the CAMS model error smaller than intra-urban variability?
Is there a „best” CAMS model for Budapest?
Is CAMS better than naive (persistence) forecasting?
Data sources
Period: 2018.12.01. – 2019.02.28.
Observed hourly PM10 data from 12 sites of the HungarianAir Quality Monitoring Network (OLM)
2% missing data
concentrations >200 μg/m3 removed (7 obs.)
1st January 0-1 UTC removed
Predicted hourly surface PM10 data of the CAMS models
0-23-hour forecasts
nearest gridpoint (values from 6 gridpoints used)
Data sources
Period: 2018.12.01. – 2019.02.28.
Observed hourly PM10 data from 12 sites of the HungarianAir Quality Monitoring Network (OLM)
2% missing data
concentrations >200 μg/m3 removed (7 obs.)
1st January 0-1 UTC removed
Predicted hourly surface PM10 data of the CAMS models
0-23-hour forecasts
nearest gridpoint (values from 6 gridpoints used)
Data sources
Period: 2018.12.01. – 2019.02.28.
Observed hourly PM10 data from 12 sites of the HungarianAir Quality Monitoring Network (OLM)
2% missing data
concentrations >200 μg/m3 removed (7 obs.)
1st January 0-1 UTC removed
Predicted hourly surface PM10 data of the CAMS models
0-23-hour forecasts
nearest gridpoint (values from 6 gridpoints used)
Winter air pollution in Budapest, 2018-2019
Histograms of hourly PM10 concentrations
Scatter plots of
hourly PM10
concentrations
Pearson correlation (incl. seasonality)
Spearman correlation (incl. seasonality)
RMSE
BIAS
Ratio of observations >50 μg/m3
CAQI accuracy
Community Air Quality Index (CAQ) PM10 concentration [μg/m3]
Very low 0-25
Low 25-50
Medium 50-75
High 75-100
Very high >100
0
0,1
0,2
0,3
0,4
0,5
Very low Low Medium High Very high
Relative frequencies of CAQI categories in Budapest during the winter 2018-2019
CAQI accuracy
Exceedance sensitivity(predicted vs. all exceedances above the 50 μg/m3 threshold)
Conclusions
SILAM forecasts were better than persistance
SILAM uncertainties were comparable to intra-urban variability
Serious underestimation of polluted events in the other models
Best CAQI accuracies and exceedance sensitivities didn’t reach 60%
Only a case study of one winter
Conclusions
SILAM forecasts were better than persistance
SILAM uncertainties were comparable to intra-urban variability
Serious underestimation of polluted events in the other models
Best CAQI accuracies and exceedance sensitivities didn’t reach 60%
Only a case study of one winter
Thank you for your attention!