Air Pollution Modelling for Support to Policy on Environment and Health Prof. David Briggs Imperial College London 3 rd GMES Forum Athens, 5-6 June 2003
Health risk and environmental impact assessment HRA and EIA are essential parts of policy they show whether intervention is needed and what effects it might have. Policies may be broad-scale, but impacts are local, so HRA and EIA are highly demanding of information. Key requirements are air quality data to identify patterns and trends in pollution. Emissions data are also needed to link source activity to effects and to identify where to intervene. But many other types of information are also needed, including population/habitats, emission sources, source activity etc.
Air Quality and Emission Data in the EU Source attribution Consistency Timeliness Long run of data Short averaging times High resolution Spatially representative Requirements ***** **** *** *** * ***** *** Air quality data bases Airbase ** ** *** ** ** ** EMEP ** * *** ** *** *** Emissions inventories CORINAIR *** ** * * *** **** EMEP emissions *** ** * **** *** ****
APMoSPHERE Air Pollution Modelling for Support to Policy on Health and Environmental Risk in Europe The Products Enhanced EU data sets on air quality (Airbase) High resolution (ca. 1km) emissions inventories (by source) High resolution (ca. 1km) air quality maps, annually updated Air quality indicator set for the EU Analysis of air quality situation in the EU and guidance on enhanced monitoring needs Assessment of EO monitoring needs
Current Data Availability Airbase NO 2
100 80 60 40 Background Industrial Traffic Unknown Airbase PM 10 monitoring sites by country 20 0 AT DE ES FI IT NL NO PT SE UK Site type 100 Unknown Rural Suburban Urban Location 80 60 40 Based on Airbase Site and Statistics data, Feb 2003 20 0 AT DE ES FI IT NL NO PT SE UK
Air Pollution Information: the Problem Ground-level air pollution is highly variably, over both time (hours-days) and space (length-scales of 100 m +) Ground-based air pollution monitoring is expensive, so can never resolve these variations adequately, and cannot therefore meet the needs for information in full Commitment of member states to provide best available information has not yet been assured so EU data sources are not optimal Existing sources of data on emissions are too coarse to provide reliable information for source attribution or air quality modelling EO data are little used and technologies are not yet able to provide measurements of pollutants of interest
Satellite data CORINE land cover etc Emissions Monitoring data CORINAIR data (NUTS3 ) Thermal imagery point sources Night-time imagery light emissions Envisat - transmissivity etc Emission maps Disaggregation Interpolation/modelling AIRBASE Air pollution maps Validation The APMoSPHERE Solution Indicators Risk assessment
Emissions modelling: the APMoSPHERE approach Level 1: national emissions totals Level 2: regionally disaggregated Level 3: locally disaggregated UK Total: 276 k tonnes County Northamptonshire Nottinghamshire Oxfordshire Powys Somerset Sussex, East Sussex, West Energy consumption 1091 1498 1782 733 1037 2158 1276 Regional statistics Light emissions Land cover EO data etc
Disaggregation of National Emissions Totals National Emissions Totals Snap 1 Combustion in Energy & Transformation Snap 2 Non-Industrial Combustion Snap 3 Combustion in Manufacturing Snap 4 Production Processes Snap 5 Extraction & Distribution of Fossil Fuels Snap 6 Solvent Use (not relevant) Snap 7 Road Transport Snap 8 Other Mobile Sources Snap 9 Waste Treatment & Disposal Snap 10 Agriculture Snap 11 Other Sources (not relevant) Employment Statistics (energy) Energy Consumption Statistics Employment Statistics Population Employment Statistics (manufacturing) Population Employment Statistics Population Employment Statistics Traffic Volumes Traffic Speeds Employment Statistics Livestock Statistics Regional Data sets NUTS 2/3 Large Point Sources (power plants) Industrial Land Cover Class * Light Emissions Industrial & Urban Land Cover Class * Light Emissions Industrial Land Cover Class * Light Emissions Industrial Land Cover Class * Light Emissions Large Point Sources (platforms) Mineral Extraction Land Cover Class Road Density By Type ** Point Locations & Activity (airports, ports) Industrial Land Cover Class Transport Networks (rail) Point Locations (disposal sites) Dumps Land Cover Class Agriculture Land Cover Class Local Data sets 1 km 2 / 200 m 2 * Weighted by Employment Statistics; ** Weighted by Traffic Volume and Speed
Preliminary results: NO 2 emissions from road traffic, Italy NB. High resolution roads data for EU are not readily available in a standard form; road traffic data are basically lacking
Level 3 Level 4 VOC emissions, England and Wales
Pollution Mapping: the APMoSPHERE Approach Kriging/co-kriging Regression mapping Bayesian hierarchical modelling Spatial stratification Maps produced for PM 10, NO 2, SO 2, CO, O 3 for mean annual concentration, mean summer and winter concentration and all Air Quality Directive parameters. Maps are developed by modelling association between environment, source activity/emissions data and measured air pollution at monitoring sites, then using these associations to predict pollution at unmonitored locations. Maps can be updated annually, as new pollution data become available, or in longer term as source activity, emissions and environmental conditions change.
Regression Mapping: Kirklees, UK NO 2 = f (traffic vol) + (built-up land) + altitude Association derived from regression modelling of onitoring sites; validated against independent monitoring sites 70 60 50 40 30 20 10 10 20 30 40 50 60 70 Huddersfield Amsterdam Prague Monitoring sites (used in model development)
Estimated annual mean background PM 10 concentration, 2001 (ug ug/m 3 ) Background concentrations are modelled using an empirically derived statistical model, calibrated against monitoring sites. Source: NETCEN
Source Activity REGIO etc Emission Totals EMEP CORINE Land Cover Airbase National/regional monitoring networks Disaggregate Data collation and checking Population, Biotopes 1km 2 Modelled Emissions by Land Class Topography, Altitude, Population, Transport, Meteorology, Light Emissions etc Enhanced EU Air Pollution Data Cluster analysis, control zone analysis Reiteration Stratification Analysis of variance 1km 2 EU-wide Air Pollution Maps Exposure Assessment at-risk populations, areas Validation Affinity Zone Mapping: the Stratification Approach
An example of affinity zone mapping: ITE land classes in GB Land classes based on a physiographic and locational variables, including altitude, temperature, rainfall, distance from sea, geology etc. Used as a basis for stratifying field surveys and deriving estimates of countryside character and condition at national and regional level.
Applications Spatially consistent and annually updated maps of air pollution for: Identification of air pollution hotspots (Air Quality Directive) Air pollution mapping in EU (e.g. SoE reports) Indicator development (e.g. EEA and national indicators) Analysis of pressures on habitats (Habitats Directive) EIA and SEA Exposure assessment (e.g. large population epidemiological studies) Health risk assessment
Another Solution Model Linkage: the HEARTS Project GUI Interpretation (e.g. re standards/health effects) Mapping/reporting Indicator construction Statistical analysis/uncertainty analysis/map smoothing Health risk assessment (dose-response models, RR) GIS (Internal) Exposure model (air pollution/noise) Group and individual level exposure assessment Journey-time, hourly and daily average exposures Intersection of T-A, population and pollution Accident/Injury model Group and individual level risk assessment Intersection of T-A, population and accident risk factors External Noise model Dispersion model Emission model Traffic model
The HEARTS Approach Pollution pattern Population distribution Integration over space Time Exposure Exposure distributions All Integration over time and space Sub-group
Preliminary Lessons from APMoSPHERE 1. EU air pollution data sets still need a lot of enhancement to be effective: requires commitment and resourcing from member states 2. Other key data sets needed to derive information products (e.g. source activity, infrastructure, population, habitats) are not fully available or difficult to acquire: EU services and data distributors are not blameless 3. The science needed to support information production and interpretation needs to be enhanced (e.g. doseresponse, sensitivity) 4. EO data promises much but has not yet delivered
Preliminary Conclusions 1. No single technology can meet the needs for information on air quality. Solutions lie in extracting information more effectively from the available (and emerging) sources by: Data linkage ground-based, EO, emissions etc Model linkage spatial models; source-to-effect 2. Key requirements are thus: Improved recognition of information needs (and differences in those needs for different applications) Increased awareness of limitations of existing monitoring and action to improve it Improved standards for data inter-comparison, exchange and linkage Improved commitment of and collaboration between institutions involved