Developing  a  hybrid  exposure  model  in  London  

Sean  Beevers  King’s  College  London  

Traffic  Pollu8on  and  Health  in  London  project    

http://www.erg.kcl.ac.uk/ResearchProjects/Traffic/Default.aspx?DeptID=ResearchProjects&CategoryID=ResearchProjectsTraffic

WP2  modelling  

MRC-HPA Centre for Environment and Health

Imperial College London

Uses:  Conges8on  charging,  Low  Emissions  Zones,  Mayor's  Air  Quality  Strategy    (hMp://www.london.gov.uk/sites/default/files/Air%20Quality%20Strategy%20v3.pdf  )  Epidemiological  research  and  Health  Impact  assessments    

London  air  quality  model  results  -­‐  hMp://data.london.gov.uk/laei-­‐2008  

Sources  Road  transport  Avia8on  sources  Passenger  and  commercial  shipping  Passenger  and  freight  rail  Large  regulated  industrial  processes  (Part  A)  Small  regulated  industrial  processes  (Part  B)  Gas  hea8ng  (domes8c  and  industrial-­‐commercial)  Oil  combus8on  sources  (domes8c  and  large  boiler  plant),  Coal  combus8on  sources  (domes8c  and  commercial)  Construc8on  source  (NRMM,  construc8on/demoli8on)  Others  (agricultural,  landfill,  waste  transfer,  accidental  fires  and  household  sources)  Biomass  burning  

Present:  Sta8c  approach  to  exposure  assessment  

Future:  Hybrid  approach  to  exposure  assessment  

Sta8c  and  Hybrid  approaches  to  exposure  assessment  

Hybrid  modelling  for  “Traffic”  

Coupled  local/regional  scale  modelling  (CMAQ-­‐urban)  

Model  outputs:  NOX,  NO2,  O3,  PM  components    Compu8ng  facili8es…..you  can’t  have  enough    Emissions  inventories  UK  Na8onal  Atmospheric  Emissions  Inventory  (NAEI)    King’s  Great  Britain  road  traffic  emissions    European  Monitoring  and  Evalua8on  Programme  (EMEP,  hMp://www.ceip.at/)    European  Pollutant  Release  and  Transfer  Register  (EPRTR)  Biogenic  Emission  Inventory  System  (BEIS  v3.14)  VOC  and  soil  NO  Eclipse  -­‐  IIASA    

Boundary  condi8ons      

Beevers  SD  ,  Kitwiroon  N,  Williams  ML,  Carslaw  DC.  2012.  One  way  coupling  of  CMAQ  and  a  road  source  dispersion  model  for  fine  scale  air  pollu8on  predic8ons.  Atmospheric  Environment  59,  pp  47-­‐58  

Hybrid  exposure  model  CMAQ-­‐urban  air  quality  model  

Micro-­‐environmental  modelling:  in-­‐vehicle  (bus,  car,  train,  tube),  cycle,  walk,  indoors  (I/O  exchange  -­‐  J  Taylor  (UCL))  

Oyster  card  data   Detailed  human  exposure  

London  Travel  Demand  Survey:  Trips  by  transport  mode:  Sex,  age,  gender  and  socio-­‐economic  status  

J  Taylor  et  al.,  2014.  The  modifying  effect  of  the  building  envelope  on  popula8on  exposure  to  PM2.5  from  outdoor  sources.  Indoor  Air.  doi:10.1111/ina.12116    

                         

 

Using the LTDS however, we have data on 85,000 people journeys for 24 hours

2005/2006 5008 households, 11583 people, 29,797 trips, 61,542 stages 2006/2007 8005 households, 18241 people, 47,029 trips, 95,930 stages 2007/2008 7873 households, 17926 people, 44,828 trips, 91,967 stages 2008/2009 8134 households, 18975 people, 43,076 trips, 89,701 stages 2009/2010 8290 households, 19187 people, 43,475 trips, 92,121 stages 2010/2011 Will get data soon 2011/2012 Will get data soon

Scaled up to the London population ~ 7 million people

The  London  Travel  Demand  Survey  (LTDS)  from  TfL  –  it’s  anonymised    

MRC-HPA Centre for Environment and Health Imperial College London

London  journeys  

Distance  travelled  by  popula8on  age  

~  25%  of  people  stay  at  home  Walk:  164,368,  Car:  123,043  Bus:  103,554,  Underground:  11,277  

PM2.5  exposure  if  you  are  65  years  and  older    (most  important  mode)  

Geographic  dose  to  NOX  (top)  and  PM2.5  (boMom)  Static model Hybrid model

NOX  and  PM2.5  Dose  by  age  group  

add some of the uncertainty being undertaken by Christina

In-­‐car  PM2.5  results  (sensi8vity)  

Model  evalua8on  

CMAQ-­‐urban  model  evalua8on  No  sites  

NO2   96  NOx   96  O3   41  PM10   106  PM2.5   68  

Fuller  GW  et  al,  2013.  Atmospheric  Environment  68,    295–296.  Fuller  GW  et  al,  2014.  Contribu8on  of  wood  burning  to  PM10  in  London.  Atmospheric  Environment  (accepted,  publica8on  pending).  Frank  N  2006  Retained  Nitrate,  Hydrated  Sulfates,  and  Carbonaceous  Mass  in  Federal  Reference  Method  Fine  Par8culate  MaMer  for  Six  Eastern  U.S.  Ci8es.  ISSN  1047-­‐3289  J.  Air  &  Waste  Manage.  Assoc.  56:500–511  

SOA  ~  1ug  m-­‐3  (Derwent  and  c.f.  observa8ons)  Biomass  burning  ~  1ug  m-­‐3  (Fuller  2013/4)  Strongly  bound  water  -­‐  (Franks  2006)  

Exposure  measurements  hMp://www.londonair.org.uk/london/asp/news.asp?NewsId=CityAirpressrelease    

Collecting other person monitoring data - Schools Study, COPE, measuring in the tube, have lots more to do.

Uncertainty  Analysis  in  Atmospheric  modelling  Quan8fying  model  uncertainty  becomes  prohibi8ve  for  models  with  long  run  8mes.    Reduce  the  number  of  parameters  through    sensi8vity  analysis  (214  to  31)    Our  solu8on  is  to  build  an  ‘emulator’          

Model  input  

Outpu

t  

Training  run  

Bayesian  Calibra?on    Each  Monte  Carlo  run  is  weighted  according  to  its  probability  of  corresponding  to  observa8onal  data    This  shivs  the  mean  of  the  uncertainty  distribu8on  towards  the  observed  value  and  reduces  its  standard  devia8on      

The  future….  

Indoor  model  -­‐  important  addi8on  to  the  current  hybrid  model    Model  evalua?on  -­‐  MRC  funded  project,  Characterising  COPD  Exacerba8ons  using  Environmental  Exposure  Modelling  (COPE)  project    Uncertainty  of  the  hybrid  approach    Develop  sources  within  the  model  that  are  currently  poorly  understood  -­‐  non-­‐exhaust  emissions,  biomass  burning,  cooking  etc      

 Thanks  for  your  aMen8on…    Thanks  to  colleagues  ERG  modelling:    NuMhida  Kitwiroon,  David  Dajnak,  Andrew  Beddows,  Chris8na  Mitsakou,  James  Smith,  Gregor  Stewart  and  Jonathon  Taylor  (UCL)    Transport  for  London  and  the  Greater  London  Authority          

   We  thank  the  Natural  Environment  Research  Council,  Medical  Research  Council,  Economic  and  Social  Research  Council,  Department  of  Environment,  Food  and  Rural  Affairs  and  Department  of  Health  for  the  funding  received  for  the  Traffic  Pollu8on  and  Health  in  London  project  (NE/I008039/1),  funded  through  the  Environmental  Exposures  and  Health  Ini8a8ve  (EEHI).  The  research  was  also  supported  by  the  Na8onal  Ins8tute  for  Health  Research  (NIHR)  Biomedical  Research  Centre  based  at  Guy’s  and  St  Thomas’  NHS  Founda8on  Trust  and  King’s  College  London.