Integrating seasonal forecasts for health impacts in ...charlie/ccrg.d/talks.d/session1.d/ccrg... · Integrating seasonal forecasts for health impacts in Africa ... ROC Skill Scores

RGS-IBG CCRG Africa April 2008 – Seasonal Forecasts and Health - Andy Morse

Integrating seasonal forecasts for health impacts in Africa – the story so far

Andy Morse, Department of Geography,

University of Liverpool

[email protected]

Acknowledgements to Anne Jones


“Our planet is filled with marvelous science-based opportunities for improving human welfare at a tiny cost,

but these opportunities are often unrecognized by policymakers and the public.”

Jeffery Sachs, Director, Earth Institute at Columbia University

writing about Neglected Tropical Diseases in Scientific American

A thought


Talk Themes

• Introduction• Background• Research Examples• Discussion & Not Conclusions – Ways Ahead


Ensemble prediction systems

EU FP5 DEMETER – seasonal ‘end-to-end’ in practice EU FP6 ENSEMBLES – s2d, ACC (AOGCM, ESSM, RCM) – towards seamless ideas and user challengesEU FP6 and NERC-UK AMMA –observation, user validation, model development, model applications EPS, trainingTHORPEX & THORPEX-Africa out to 15 days

Introduction -Project Links and Roles

http://wcrp.wmo.int/

http://www.clivar.org/index.php

http://www.royalsoc.ac.uk/

http://www.ensembles-eu.org/

http://www.nerc.ac.uk/

http://www.pa.op.dlr.de/stiss/


Introduction

• Important connections between climate and disease

• Climate variability can be important for epidemics

• Climate is not the only factor in causing diseases – even those with strong climate drivers

• Limited understanding of climate by health practitioners (& vice versa) – unlike climate and agriculture.

• Significant challenges to get decision and policy makers to useclimate information – in development and health

• Chance to develop early warning systems to improve preparednessand targeting of meagre resource in area with poor health services


IntroductionChallenges to improving climate - health research

• Interdisciplinary – language – working and research practises

• Lack of contact – academic risks – lacking cross disciplinary funding

• Lack of knowledge (and health tailored dissemination) of forecasts/climate products

• Uncertainty in climate forecasts, access and use of climate data

• Health data – rarely integrated, paucity, quality and access

• Critical mass of researchers


IntroductionWeather and Climate Models

• Numerical weather forecasting – single high resolution model (25km) few days

• Medium range ensemble prediction systems (EPS) 10 to 15 days (80km) 50 members

• Monthly EPS - ‘just available’ – persisted SST

• Seasonal EPS 180 day integrations 50 members (125km) coupled ocean

• Decadal scale EPS very experimental – currently 13 months out to 10 years

‘decadal gap’ period 2010 to 2050

• Climate models – typically run through late 20th century out to 2100 (100 to 200km)multiple single model runs - range of scenarios


Introduction

IPCC 2007


Background

Source: UNEP GRID Arendalhttp://www.grida.no/


Background

Source: UNEP GRID Arendal http://www.grida.no/


Background

WorldMapper – Peters equal area projection

Source: http://www.worldmapper.org© Copyright 2006 SASI Group (University of Sheffield) and Mark Newman (University of Michigan)


Background

WorldMapper – malaria deaths



Background

WorldMapper – killed by drought



Background - malaria

Mean monthly climatic data at Bukoba(a) and malaria cases reported by Ndolage hospital (b) for the period 1991–1999.

Jones et al. Malaria Journal 2007 6:162 doi:10.1186/1475-2875-6-162

http://www.malariajournal.com/content/6/1/162/figure/F2?highres=y


Background

• Time between trigger threshold to epidemic peak often too short to take effective

intervention – need for skilful and timely seasonal climate forecast

Epidemic Cycle

0

2 04 0

6 0

8 0

100120

140

9745

9748

9751

9802

9806

9809

9812

9815

9818

9821

9824

9827

9830

9833

9841

9844

9847

R ep o r ting w e ek

Nu

mb

er o

f ca

Vaccine

ThresholdEffect


Background – ensemble prediction systems – seasonal forecasts

Chart from ECMWF


Background – EPS seasonal

Chart from ECMWF

Sahel


Background – EPS medium range

Chart from ECMWF


Banizoumbou AWS Jul-2 to Sept-1 rainfall

0

5

10

15

20

25

30

35

40

4518

3

186

189

192

195

198

201

204

207

210

213

216

219

222

225

228

231

234

237

240

243

Julian Day 2006

Rai

n (m

m)

Background – rainfall – single season


Background – rainfall interannual

Niamey-Aero, Niger 13.5N 2.1E

0.0

200.0

400.0

600.0

800.0

1000.0

1200.0

1900 1920 1940 1960 1980 2000

Ann

ual R

ainf

all (

mm

)


Background –real work


The Forecasting Triangle

DemandForecasts

Training + Product Guidance and Development

DisseminationDissemination

FeedbackFeedback

Providers Users

Developers with users and providers

Morse in prep.


Background -scalesGlobal model – regional impacts – local and microscale processes

kms to 100s m cm to mm

1000s to 100s km metre

Africa to mosquito 9 orders of magnitudeEarth-Sun distance to

galaxy scale


Research Examples – verification paradigm

from Morse et al. (2005)Tellus A 57 (3) 464-475


Research Examples – verification

Original MARA map Craig et al., 1999

www.mara.org.za

Based on model Craig et al. 1999 www.mara.org.zarun with ERA-40

slide from Anne Jones, University of Liverpool


-0.42 -0.35 -0.28 -0.21 -0.14 -0.07 0.07 0.14 0.21 0.28 0.35 0.42

Research Examples – predicting rain anomalies

-0.42 -0.35 -0.28 -0.21 -0.14 -0.07 0.07 0.14 0.21 0.28 0.35 0.42

NDJF DEMETER ensemble mean precipitation anomaly (mm/day) for i) five highest malaria years, ii) five lowest malaria years in Botswana

from M.C. Thomson, F.J. Doblas-Reyes, S.J. Mason, R. Hagedorn, S.J. Connor, T. Phindela, A.P. Morse, and T.N. Palmer (2006). Malaria early warnings based on seasonal climate forecasts from multi-model ensembles, Nature, 439, 576-579.


Research Examples – statistical models

Quadratic malaria relationship from Thomson et al. (2005) Malaria Index for Botswana (1982 to 2002)


Research Examples – dynamical modelsbiting/laying:

temperature dependent

sporogoniccycle:

temperature dependent

larval stage:

rainfall dependent

After CDC etc.


Research Examples – malaria modelling

TemperatureTemperature

Mosquito survivalafter Martens (1995) slide from Anne Jones unpublished Ph.D. thesis

At T = 25°C sporogonic cycle length = 15.9 days

2.9% survive to infectious stage


00.05

0.10.15

0.20.25

0.30.35

0.40.45

1 31 61 91 121 151

Forecast Day

Mal

aria

Pre

vale

nce

Research Examples – malaria prediction plume

95

85

65

35

15

5ERA

Botswana malaria forecast for February 1989, LMM driven by DEMETER multi-model

(ERA-driven model shown in red)

Plot from Anne Jones unpublished Ph.D. thesis University of Liverpool


Research Examples – malaria modellingTier-2 malaria runs - ROC Skill Scores Above Median Event

Nov 2-4 Nov 4-6

DEMETER data set. Areas of high interannual variability were selected and persisted forecast skill was removed from the scores.

Jones, A. and Morse, A. (2007) CLIVAR Exchanges, 43


Fig. 2: (A) Differences in the annual average model prevalence (in %) and (B) in the standard deviation regarding the annual maximum of the model prevalence (in %) between the last decade of the A1B scenario (2041-2050) and the past period (1960-2000).

Changes in the malaria distribution University of Liverpool, A. Morse & A. JonesUniversity of Cologne, V. Ermert & A. FinkUniversity of Würzburg, H.Paeth

LMM malaria scenarios (2041-2050):• decreased malaria transmission due to precipitation reduction• reduced model prevalence variability in N-Sahel ⇒ fewer epidemics/malaria retreat• 13-16°N: increased variability in the S-Sahelian zone ⇒ more frequent epidemics

in denser populated areas• farther south: malaria transmission remains stable


• Allow non-linear mapping of combined ensemble PDFs through time • Allow assessment of downscaling, dressing of ensembles etc.• Define forecast skill and potential user/societal value• Make link to decision makers/stakeholders• Allow linkage across modelling streams – semi seamless approach• Allow assessment of skill improvement across model cycles.

• African users – clear forecasting needs for rains – onset, break cycles, cessation – intraseasonal and interseaonal – early warning of high impacts events

Discussion - Climate Impacts – Integration of users


Not Conclusions – Ways Ahead

Increasing interest in climate-health links particularly with operational predictions-EPS at medium range, seasonal and climate scales

Need to undertake underpinning health science and integrated surveillance

Need to raise awareness at all levels – students and practitioners to researchers to decision and policy makers

Need to build wider community – few clinicians & further links to zoonoses etc.

Education and training - public and health community and climate community

Funding for short term embedding in climate groups, short courses and pilot projects


Questions



DEMETER, ENSEMBLES, AMMA, THORPEX, CLIVAR

Integration impacts models – Ensemble Prediction SystemsProbabilistic – all lead timesPost processing – downscaling

Continuum: forecast model to customerInterdisciplinary – networking – cross cuttingTimely use of existing climate information


User driven – tailoring product, skill requirements, ‘acceptable’ uncertaintyScience – seamless approach, impact models, downscaling, risks, feedback model development, adaptationPolicy – decisions to impact reductionTechnical – ensembles, data, cross cutting, model climates, mitigationTraining – probabilistic – use, validation & uncertainty

Integrating seasonal forecasts for health impacts in ...charlie/ccrg.d/talks.d/session1.d/ccrg... · Integrating seasonal forecasts for health impacts in Africa ... ROC Skill Scores

Documents