1) Climate services aim to manage risks from climate variability and change by developing relevant tools from climate information. This includes partnerships between climate and health professionals to create early warning systems for infectious diseases. 2) Seasonal climate forecasts can provide warnings for disease outbreaks months in advance. A dengue forecasting model developed for Brazil uses climate and epidemiological data to predict dengue risk at a local level. 3) Climate forecasts were used to predict increased dengue risk in Machala, Ecuador in 2016 due to anticipated heavy rainfall and warm temperatures from El Niño. The outbreak peaked earlier than expected according to the forecast.

1. Climate services and early warning systems
for infectious disease outbreaks
Rachel Lowe
Royal Society Dorothy Hodgkin Fellow
London School of Hygiene & Tropical Medicine
24 October 2017, THET annual conference, London, UK

2. Global framework for climate services (GFCS)
2
• Manage the risks of climate variability and change.
• Transform climate information into relevant, usable decision-support tools.
All these sectors have a big impact on our health!

3. Climate services for health
• Partnership
• Research
• Product development & delivery
• Evaluation
• Capacity building
• Co-developed by health and climate
professionals.
3
Lowe et al. Dengue EWS
Lowe et al. Capacity Building

4. From global observations to local interventions
4
The Economist

5. Early warning and response systems
• Early warning systems that account for
multiple disease risk factors can help to
implement timely control measures.
• Seasonal climate forecasts provide an
opportunity to anticipate epidemics
several months in advance.
5

6. Dengue in Brazil
6
• Model framework developed in collaboration with
European-Brazilian climate and health institutions.
• Data (dengue, climate, cartographic, demographic,
socio-economic) to formulate model, produce
probabilistic dengue predictions for >550 microregions.
• Optimum trigger alert thresholds determined for
scenarios of medium-risk and high-risk of dengue,
according to incidence alert levels defined by the
Ministry of Health.

7. Lowe et al., 2014, Lancet Infect Dis
Probabilistic dengue forecasts in Brazil
7
• Early warning framework applied to predict dengue risk for the World Cup in Brazil.
• Category boundaries: 100 and 300 cases per 100,000 inhabitants.

8. Probability of observing correct category
8
Low Medium High
Lowe et al., 2016, eLife

9. Comparison of forecast to null model
9
Comparison of hit rate and false alarm rate for forecast model (blue) and
seasonal average null model (orange) for June 2000-2014.
2014 event
hit rate: 57% (33%)
false alarm rate (type I error rate): 23% (13%)
miss rate (type II error rate): 43% (67%)Lowe et al., 2016, eLife

10. El Niño and dengue in Ecuador
10
• El Niño a robust predictor of dengue
outbreaks in El Oro province.
• Seasonal climate forecasts more skillful
during El Niño events.
• Forecasts of temperature, rainfall and El Niño
could provide dengue early warnings.
Stewart-Ibarra & Lowe, 2013, AJTMH
El Niño
parameter
estimate
El Oro
El Niño and rainfall association (JFM, 2 mon lag)

11. 11
On February 26, 2016, over 170 mm of rain fell in 10 hours, and coincided
with high tides, causing the worst flooding since the 1997-1998 El Niño.

12. Climate and dengue associations
12
Cooler and drier than usual  less dengue
Warmer and wetter than usual  more dengue
Lowe et al., 2017 Lancet Planet Health
Precipitation Temperature
Ensemble climate forecasts (24 members) for
precipitation and minimum temperature 2016, Ecuador
DENGUE PRECIP
MIN TEMP ENSO

13. Current practice: average over last 5 years
13

14. Climate-driven dengue forecast, Machala 2016
14
84% chance of
exceeding
threshold of
95% upper CI
for previous
five years
Month
log(dengueincidencerate)
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
0246810
J F M A M J J A S O N
●
●
●
Mean (2011−2015)
Upper 95% CI (2011−2015)
Predicted median 2016
95% prediction interval

15. Peak occurred earlier than expected
15
Month
log(dengueincidencerate)
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
0246810
J F M A M J J A S O N
●
●
●
●
Observed
Mean (2011−2015)
Upper 95% CI (2011−2015)
Predicted median 2016
95% prediction interval
Lowe et al., 2017 Lancet Planet Health
Timing: climate forecasts
Magnitude: correct misreporting

16. Decision making scenarios
16
LATIN AMERICA
QUEENSLAND
MEDITERRANEAN
UK
• Scenario A:
Manage simultaneous outbreaks
in endemic settings
(dengue, chikungunya, Zika, yellow fever).
• Scenario B:
Resist local transmission
in disease free yet mosquito-infested areas
(infected returned travellers, suitable climate).
16

17. Thank you for your attention!
Questions?
rachel.lowe@lshtm.ac.uk
@drrachellowe
17