A global assessment on temperature, climate and health · 2017. 12. 11. · Daniel Oudin Astrom Martina Ragettli Yue-Liang Leon Guo Chang-fu Wu Michelle Bell Antonella Zanobetti Joel

Intro MCC Models Projections Projections Discussion

A global assessment on temperature,climate and health

Results from the MCC Project

Antonio Gasparrini

London School of Hygiene & Tropical Medicine, UK

Joint Research Centre (JRC)Ispra, Italy, 4–6 October 2017

Gasparrini A LSHTM

A global assessment on temperature, climate and health


Outline

1 Epidemiological research on temperature

2 The MCC Network

3 Modelling framework

4 Future impacts: RCPs

5 Future impacts: Paris Agreement

6 Discussion

Gasparrini A LSHTM



July/August 2003 heat wave in Europe

[From NASA Earth Observatory]

Gasparrini A LSHTM



Impact on mortality in Paris

2000 2002 2004 2006 2008 2010

200

400

600

800

Daily mortality in Paris: 2000−2010 and summer 2003N

umbe

r of

dea

ths

Jun Jul Aug Sep Oct

200

400

600

800

Num

ber

of d

eath

s

Gasparrini A LSHTM



Climate change: the ’hockey stick’

[Adapted from Mann et al, Geo Res Lett (1999)]

Gasparrini A LSHTM



Limitations

Complexities in modelling temperature-health relationships:non-linearity, lagged effects, pooling multi-parameter estimates

Geographical heterogeneity: little information on susceptibility factorsresponsible for modifying the risk, difficult to compare estimates fromdifferent areas

Poor knowledge about key aspects of the relationship, such asadaptation and acclimatization

Limited statistical power in location-specific assessments

Gasparrini A LSHTM



The MCC Network

The Multi-City Multy-Country (MCC) Research Network is an internationalcollaboration aiming at producing epidemiological evidence on associationsbetween weather and health

Advantages:

Global perspective: data from multiple locations within severalcountries, including populations with different characteristics and exposedto various climatic conditions

Flexible modelling framework allowing non-linear/lagged responses,separation of effects due to cold/heat and moderate/extremetemperature, and assessment of effect modification

Gasparrini A LSHTM



MCC participants and funding

Participants:

Antonio Gasparrini

Ana Maria Vicedo-Cabrera

Francesco Sera

Ben Armstrong

Yuming Guo

Shilu Tong

Micheline Stagliorio Coelho

Paulo Saldiva

Eric Lavigne

Patricia Matus Correa

Nicolas Valdes Ortega

Haidong Kan

Samuel Osorio

Jan Kysely

Ales Urban

Jouni Jaakkola

Niilo Ryti

Mathilde Pascal

Ariana Zeka

Patrick Goodman

Paola Michelozzi

Matteo Scortichini

Masahiro Hashizume

Yasushi Honda

Magali Hurtado

Julio Cesar De La Cruz

Xerxes Seposo

Noah Scovronick

Fiorella Acquaotta

Ho Kim

Aurelio Tobias

Carmen Iniguez

Bertil Forsberg

Daniel Oudin Astrom

Martina Ragettli

Yue-Liang Leon Guo

Chang-fu Wu

Michelle Bell

Antonella Zanobetti

Joel Schwartz

Tran Ngoc Dang

Do Van Dung

Funding: Medical Research Council UK [G1002296; MR/M022625/1]

Gasparrini A LSHTM



The MCC dataset

Largest dataset ever collected: data from 512 locations in 26 countries withinthe period 1972–2012, including 105 million deaths

Countries: Australia, Brazil, Canada, Chile, China, Colombia, Czech Republic,Finland, France, Iran, Ireland, Italy, Japan, Mexico, Moldova, Philippines, SouthAfrica, South Korea, Spain, Sweden, Taiwan, Thailand, UK, USA, Vietnam

Daily time series of mortality counts (all-cause, CVD, respiratory, others),temperature (mean, min, max), humidity, air pollution (PMx, O3, NOx, SO2,CO)

Location-specific meta-variables on climatological, geographical,infrastructural, demographic, socio-economic characteristics

Gasparrini A LSHTM



Map of MCC locations

●

●

●●

●

●●●

●

●

●

●●

●●

●

●●

●

●

●

●

●● ●●●●

●●●●●●●●●●●●●●

●●●●

●●●●●●●●

●●●●●●

●●

●●

●●●●●

●●

●●

●●

●●

●

●

●●●

●

●

●●●

●

●

●

●

●●

●●

●●●●

●●●●●●

●●●●

●●●●●●●●●

●●● ●●

●

●●●●●●

●●●●

●●●●●●

●●

●●●●● ●

●●● ●●●●●●●

●●● ●●●

●●●●●●●●●●●●●●

●

● ●●●

● ●●●●●●

●

●●●●●

●●

●

●●●● ●●●●●●●●

●

●●

●

●●

●●

●

●●

●●●●

●●

●●

●●

●

●

●●●

●

●●●

●●●●

●

●●●

●● ●●●●

●●●●

●●

●●●

●●

●●● ●●●

●●●●●●

●

●● ●

●●●●●●●

●●●●●●●●●●

●●●●●●●●●●●●

●●●●

●●●●●●●● ●●

●●●●

●●●●

●●●● ●●●●●●●●●●●●

●●●●●●●

●●

●●

●●●●●●●●●●

●●●●

●●●●

●●

●●●●●●●●●● ●●

●●●●●

●●●●●●●●

●●●

●●●●●●●

●●

●●

●●●

●

●

●●●●●●

●●●

●

●

●

●●●●

●

●●●●●●●●●●●●●

●●●●

●

●●●

●●

●

●●●

●

●

●●●●●●●●●●●●●●●●●●●●

●●●

●●

●●

●●

●●

●

●

●

●

●

●●●●●

●●

●

●

●

●

●

●

● ●● ●

●

●●

●●●

●

●●

●●

●●● ●●

●

●

●

●

●●●

●

●●

● ●●●

●

●

● ●

●

●

●●●

●

●

●

●

●

●●●●●●●

●

●

●●

●●●

●

●

●●● ●

●

●●

●

●●●

●

●●

●●

●●

●

●

●

●

●●

●●

●●

●

●

● ●

●●

●

●●●

●

● ●●●●●

●

●

●●

●●

●

●

●

●

●●

●

●

●

●●●●

●

●

●

●

●

●● ●●

●

●●

●

●

●

●

●●●

● ●

●●●●●

●●

● ●●

●●●

●

●

●

●

●●

●

●●

●

●

●●●●● ●

●●

●

●

●

●●

●

●●●

●● ●

●●●

●●●

●

●

●

●

●

●●

●

●

●

●●

●●●

●●

0 2000 4000 km

●●●●●●●●

AverageTemperature (°C)

28C

Gasparrini A LSHTM



Complex temperature-health relationships

[From Curriero et al, AJE (2002)]

Gasparrini A LSHTM



New methodological issues

Complexities in modelling the exposure-response relationships:non-linearity, lagged effects, pooling multi-parameter estimates

Geographical heterogeneity: little information on susceptibility factorsresponsible for modifying the risk, difficult to compare estimates fromdifferent areas

Adaptation and acclimatization: evidence of temporal changes, howeverlittle info on its timescale or drivers

Gasparrini A LSHTM



Modelling framework

1 Two-stage time series analysis to derive historical temperature-mortality

relationships using observed data:

First-stage quasi-Poisson GLM with distributed lag non-linearmodel (DLNM) expressing a bi-dimensional exposure-lag-response

Second-stage multivariate meta-regression to pool estimates whileaccounting for location-specific characteristics

2 Computation of historical excess mortality for total and cold/heatcomponents in each location, then aggregated by area and period

3 Derivation of future daily temperature and mortality series projectedin the next decades

4 Projection of future health impacts and associated uncertainty

Gasparrini A LSHTM



Exposure-lag-response associations

Temperature (C) 010

2030

Lag

0

5

10

15

20

25

RR

0.9

1.0

1.1

1.2

1.3

1.4

Exposure−lag−response surfaceLondon 1993−2006

Temperature

RR

−5 0 5 10 15 20 25 30 35

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Overall cumulative exposure−response curveLondon 1993−2006

Gasparrini A LSHTM



ISI-MIP dataset

Global simulation dataset of daily meteorological variables includinghistorical (1860–2005) and projection (2006–2100) periods

Simulations from 5 global climate models (GCM) (HadGEM2-ES,IPSL-CM5A-LR, MIROC-ESM-CHEM, GFDL-ESM2M, and NorESM1-M)from the CMIP5 archive of IPCC, under 4 emission scenarios (RCP2.6,RCP4.5, RCP6.0, and RCP8.5)

Downscaled over a 0.5◦ × 0.5◦ grid through bias correction usingre-analysis temperature data

Gasparrini A LSHTM



Scenarios of global warmingGasparrini et al, in submission

0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●●

●●

● ● ● ● ● ● ●

●●

●●

●●

●● ● ●

●

●● ●

● ●● ●

●●

●●

●●

●●

●

●

●

●

●●

●

North America

RCP2.6RCP4.5RCP6.0RCP8.5

0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●●

●●

● ● ● ● ● ● ●

●● ●

●●

●● ●

● ● ●

●● ●

●●

● ●●

●●

●

●●

●●

●●

●●

●●

●

Central America


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●● ●

● ●● ● ● ● ● ●

●● ●

●● ●

● ●● ● ●

●●

● ●●

● ●●

●● ●

●●

●●

●●

●●

●●

●

South America


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●●

●● ● ● ●

● ● ●●

●● ●

●●

● ●● ● ● ●

●●

●● ●

●●

●●

● ●

●● ●

● ●●

●●

●●

●

North Europe


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●●

●● ● ●

● ● ● ● ●

●● ●

●●

● ●● ● ● ●

●●

●● ●

● ●●

●● ●

●●

●

●●

●●

●

●

●●

Central Europe


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●●

●● ●

● ● ● ● ● ●

●●

●●

●● ●

● ●● ●

●●

●● ●

● ●●

●●

●

●●

●●

●●

●●

●

●

●

South Europe


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

● ●●

● ●● ● ● ● ● ●

● ●●

●● ●

●● ● ● ●

● ●●

● ●● ●

●

●●

●

● ●●

●●

●●

●

●

●●

East Asia


0

2

4

6

Year

Incr

ease

in te

mpe

ratu

re (

°C)

1990−99 2020−29 2050−59 2080−89

0

2

4

6

●● ●

● ●● ● ● ● ● ●

●●

● ●● ●

● ●● ● ●

●●

● ●●

● ●●

●●

●

●●

●●

●●

●●

●●

●

South−East Asia


0

2

4

6

YearIn

crea

se in

tem

pera

ture

(°C

)1990−99 2020−29 2050−59 2080−89

0

2

4

6

●● ●

● ●● ● ● ● ● ●

●● ●

●● ●

● ●● ● ●

●●

● ●● ●

●● ●

● ●

●●

●●

●●

●●

●●

●

Australia


Gasparrini A LSHTM



Example with two citiesGasparrini et al, in submission

LondonR

R

0.5

1.0

1.5

2.0

2.5

3.0NorESM1−M − RCP8.5

Tem

pera

ture

dist

ribut

ion

0.00

0.02

0.04

0.06 2010−192090−99

Exc

ess

mor

talit

y (%

)

0.000

0.002

0.004

0.006

0.008

0 5 10 15 20 25 30

Temperature (°C)

Manila

RR

0.5

1.0

1.5

2.0

2.5

3.0NorESM1−M − RCP8.5

Tem

pera

ture

dist

ribut

ion

0.0

0.1

0.2

0.3 2010−192090−99

Exc

ess

mor

talit

y (%

)

0.00

0.01

0.02

0.03

0.04

0.05

24 26 28 30 32 34 36

Temperature (°C)

Gasparrini A LSHTM



Projections under RCPsGasparrini et al, in submission

Exc

ess

mor

talit

y (%

)

● ● ● ● ● ●

● ● ● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

● ● ● ● ● ●

● ● ●● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

● ● ● ●● ●

● ● ●●

●●

1990

s20

10s

2030

s20

50s

2070

s20

90s

North America161 locations

RCP2.6 RCP4.5 RCP8.5

0

5

10

15

Exc

ess

mor

talit

y (%

)

●●

● ● ● ●

● ●● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

● ●● ●

● ●● ●

● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●

●●

●● ●

● ●●

●

●

●

1990

s20

10s

2030

s20

50s

2070

s20

90s

Central America10 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●● ● ● ● ●

● ●● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

● ● ● ●● ●

●●

● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●● ● ●● ●

●●

●

●

1990

s20

10s

2030

s20

50s

2070

s20

90s

South America22 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●● ● ● ● ●

● ● ● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

● ● ● ●

● ● ●● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●●

●●

● ● ●●

●●

1990

s20

10s

2030

s20

50s

2070

s20

90s

North Europe18 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)●

● ● ● ● ●

● ●● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

● ● ● ●

● ●●

● ●●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●●

●●

● ●●

●

●

●

1990

s20

10s

2030

s20

50s

2070

s20

90s

Central Europe34 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●● ● ● ● ●

●●

● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

● ● ● ●

● ●●

●● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●●

●●

● ●

●

●

●

●

1990

s20

10s

2030

s20

50s

2070

s20

90s

South Europe63 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●● ● ● ● ●

● ● ●● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

● ●● ● ● ●

● ● ●● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●●

●●

● ● ●●

●●

1990

s20

10s

2030

s20

50s

2070

s20

90s

East Asia69 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●●

● ● ● ●●

●● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●● ● ●

●●

●

●● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●● ● ●●

●

●

●

●

●

1990

s20

10s

2030

s20

50s

2070

s20

90s

South−East Asia71 locations


0

5

10

15

Exc

ess

mor

talit

y (%

)

●● ● ● ● ●

● ● ● ● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●● ● ●

● ● ●● ● ●

1990

s20

10s

2030

s20

50s

2070

s20

90s

●●

●●

●●

● ● ●●

●●

1990

s20

10s

2030

s20

50s

2070

s20

90s

Australia3 locations


0

5

10

15

Gasparrini A LSHTM





Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15North America

161 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15Central America

10 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15South America

22 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15North Europe

18 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15Central Europe

34 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15South Europe

63 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15East Asia

69 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15


−10

−5

0

5

10

15South−East Asia

71 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

RCP2.6 RCP4.5 RCP8.5D

iffer

ence

in e

xces

s m

orta

lity

(%)

−10

−5

0

5

10

15


−10

−5

0

5

10

15Australia

3 locations

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

2010

s20

30s

2050

s20

70s

2090

s

Gasparrini A LSHTM




●

●●

●

●●●

●

●

●

●●●

●

●

●●

●

●

●

●

● ●●

●●●●● ●●●

●

●●●●

●●●

●

●

●●●●●

●●

●●

●●

●●

●

●

●●

●

●

● ●

●

●

●

●●●●

●

●●●●●●●

●●●●●●

●●● ●●

●●●●●●

●●●●●

●●●●●●

●●●●●●●●●●●●●

●

● ●●●

● ●●●●●●

●●

●●●●

●●

●

● ●●● ●●●●●●●●●

●●

●

●●

●●

●

●●

●●●●

●●

●●

●●●●●●●

●●●●●

●● ●●●

●●

●●●●●●●●

●●

●● ●●●●●●

●●●● ●

●

●●●●●●●

●●

●

●●●●●●

●

●●●●●●●●

●●●

●●●●●●●●●●

●●●●●

●

●●●●●●

●●●

●

●

●

●●●●

●

●●●●●●●●●●

●●●

●●●●

●

●●●

●●

●

●●●

●

●

●●●●●●●●●●

●●

●

●●

●●

● ●

●

●

●

●

●

●●●●●

●●●

●●

●

●

●

● ● ● ●

●

●●

●●●

●

●●

●●

●●● ●●

●

●

●

●

●●●

●

●●

● ●●●

●

●

● ●

●

●

●●●

●

●●

●

●

●●●●●●●

●

●●●

●●●

●●

●●● ●

●

●●

●

●●●

●

●●

●●

●●

●

●●

●

●●

● ●●●

●

●

● ●

●●

●

● ●●

●

● ●●●●

●●

0 2000 4000 km

●●●●●●

Net change inexcess mortality

−30 to −5%−5 to 0%0 to 5%5 to 10%10 to 15%15 to 50%

Gasparrini A LSHTM



Scenarios: 20-year windows

Gasparrini A LSHTM



Projections under Paris AgreementVicedo-Cabrera et al, in submission

Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15North America

Diff

eren

cein

exc

ess

mor

talit

y (%

)

Scen 2C Scen 3C Scen 4C

North Europe


East Asia


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15Central America

Diff

eren

cein

exc

ess

mor

talit

y (%

)


Central Europe


South−East Asia


Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−10

−5

0

5

10

15South America

Diff

eren

cein

exc

ess

mor

talit

y (%

)


South Europe


Australia


Gasparrini A LSHTM




Difference in excess mortality 2C vs 1.5C (%)

−1.0 −0.5 0.0 0.5 1.0 1.5 2.0

Australia

South−East Asia

East Asia

South Europe

Central Europe

North Europe

South America

Central America

North AmericaHeat−related EMCold−related EMNet EM

Gasparrini A LSHTM




Diff

eren

ce in

exc

ess

mor

talit

y (%

)

−15

−10

−5

0

5

10

15

Tropical

Diff

eren

cein

exc

ess

mor

talit

y (%

)

Scen

2C

Scen

3C

Scen

4C

Arid

Scen

2C

Scen

3C

Scen

4C

Temperate

Scen

2C

Scen

3C

Scen

4C

Cold

Scen

2C

Scen

3C

Scen

4C

Gasparrini A LSHTM



Discussion

This represents the largest epidemiological study on health impact projectionsunder climate change scenatior, covering hundreds of locations in variousregions

Results show a general pattern of increase in temperature-related excessmortality, especially under more extreme scenarios, but with importantgeographical differences

Impacts are much reduced under milder global warming scenarios,confirming the benefits of the implementation of mitigation policies to reduceGHG emissions

Plan to expand the data and analysis to areas not covered in the current study,such as Africa and the Middle East, to attempt a global assessment

Need to extend the analysis to less simplistic scenarios including adaptationand changes related to pathways in socio-economic, demographic,infrastractural characteristics

Gasparrini A LSHTM


Documents

A global assessment on temperature, climate and health · 2017. 12. 11. · Daniel Oudin Astrom Martina Ragettli Yue-Liang Leon Guo Chang-fu Wu Michelle Bell Antonella Zanobetti Joel