Climate/Health Modeling. Ben Zaitchik Johns Hopkins University

Size: px

Start display at page:

Download "Climate/Health Modeling. Ben Zaitchik Johns Hopkins University"

Gyles Sharp
5 years ago
Views:

1 Climate/Health Modeling Ben Zaitchik Johns Hopkins University

2 A Guide to the Perplexed (abridged) Trends vs. variability Targets Memory MIP? Movement

3 Trends vs. Variability Lyme Disease

4 Trends vs. Variability Different: decisions datasets modeling approaches uncertainties

5 Trends vs. Variability: Heat waves Heat Wave Trends: days/year Smith et al. (2014)

6 Trends vs. Variability: Heatwaves NW Regionalization on Summer Temperature Variability SW NGP TXMX SGP NE SE

7 Trends vs. Variability: Heat waves Summertime Temperature Composites -2C -1C 0 +1C +2C +3C

8 Trends vs. Variability: Vibrio Vibrio vulnificus probability Urquhart et al. (2014)

9 Trends vs. Variability: Vibrio GLM GAM Urquhart et al. (2014)

10 Targets Lead time Chain of analysis Performance metric

11 Targets: lead time Value Accuracy Lead Time

12 Targets: chain of analysis Temperature, Rainfall Soil Moisture, Vegetation Health Vector Distribution Transmission Risk

Example: Hierarchical Bayesian model of Malaria risk Water bodies Model 1: water bodies Water bodies with larva from species i Water bodies with larva from species j Water bodies with larva from

13 Example: Hierarchical Bayesian model of Malaria risk Water bodies Model 1: water bodies Water bodies with larva from species i Water bodies with larva from species j Water bodies with larva from species k Model 2: larva Adult mosquito abundance from species i Adult mosquito abundance from species j Adult mosquito abundance from species k Model 3: adult mosq. Malaria infection Model 4: infection Denis Valle, Univ of Florida

14 Targets: chain of analysis Temperature, Rainfall Soil Moisture, Vegetation Health Vector Distribution Transmission Risk What suite of information is most useful for health officials?

15 Example: the USDM

16 Targets: performance metric

17 Targets: performance metric

18 The Persistence of Memory Seasonal forecasts don t necessarily require climate forecasts

19 Memory: Infectious Disease Murray Valley Encephalitis Virus (MVEV) can be predicted using satellite observations at 2 month lead In our Amazon malaria risk model, satellite rainfall rate is most predictive at 10 week lead. Rift Valley Fever warnings in East Africa make use of satellite observations integrated over 3 months. For hantavirus, seasonal rainfall anomalies can influence cases two or more years later. Cholera dynamics have been related to precipitation at lead times from days to seasons.

20 Memory: Food Insecurity FEWSNET

21 Memory: Food Insecurity Current climate observation + past performance [operational] Current climate observation + food system models [~operational] Large scale dynamical model + statistical regression [experimental] Fully dynamical seasonal forecast [experimental]

22 To MIP or not to MIP

23 MIP MIP vs. model challenges What would a VectorMIP look like? What are the science questions?

24 And yet it moves Vectors are on the move: ecological range is often larger than observed distribution People are on the move: travel, settlement, and migration Behaviors change (humans, animals, and vectors)

25 Travel Azman, Lessler, et al. (2013) Orange D4D Submission

26 Settlement & Migration

27 Settlement & Migration Annual Parasite Index Frontier Malaria Years since settlement

28 Human Malaria Infection Climate & Hydrology

29 Climate & Hydrology Land Use Anopheles Larva Habitat Density of Anopheles Adults Human Malaria Infection

30 Infrastructure Migration, Colonization, and Agriculture Land Use Anopheles Larva Habitat Climate & Hydrology Human Exposure Density of Anopheles Adults Human Malaria Infection

31 Behavioral Change

32 # Cases (1000s) Reported Malaria, Loreto Province P. vivax P. falciparum MAJOR FLOOD * Highest Deforestation Rate in Peru Iquitos-Nauta Road Paving & Fujimori logging concessions Roll Back Malaria 2 nd highest increase in the Amazon to 2006, Major decline to 2011 Bill Pan, Duke Univ.

33 Behavioral Change Seth Guikema, U Mich

34 A Guide to the Perplexed There are opportunities for projection and prediction, but the opportunities differ Actionable information can be found throughout the analysis chain Forecast has multiple meanings It can be useful to plan a MIP, even if we don t pursue it Change matters, and we can address it

35 Thank You

Climate Change: The Public Health Response

Climate Change: The Public Health Response Climate Change: The Response Abstract There is scientific consensus that the global climate is changing, with rising surface temperatures, melting ice and snow, rising sea levels, and increasing climate