Understanding the role of land cover / land use nexus in malaria transmission

under changing socio-economic climate in Myanmar

Tatiana Loboda, Kathleen Stewart, Julie Silva (UMD)Mark Carroll (SSAI/GSFC)

Myaing Nyunt, Chris Plowe, Poe Poe Aung (DGHI)

Gething et al. 2011 – Available via Malaria Atlas Project

Fig 3 shows predictions categorized as: • low risk - light red; • intermediate risk - medium

red;• high risk - dark red.

P. falciparum

Source: pixnio.com Source: http://www.saburchill.com/ans02/chapters/chap013.html

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

Processing and analysis grid• Common grid: ½˚ longitude x

½˚ latitude projected into UTM zone 46

• Optimizes raster images at fine resolution (1 – 2 m) and moderate resolution (30 m)

VHR Panchromatic (left) and Multispectral (right) data for Rakhine

587 Multispec scenes557 Pan scenes

• Very High Resolution (VHR): WorldView @ 1 – 2 m pixel

• > 1,100 scenes over 2010 –2017

• Able to ortho-rectify and convert to TOA reflectance all images in ~12 hours on 10 virtual machines

Sample mosaic in Rakhine before (left) and after (right) adjustment

• A geolocation problem (a mismatch in the geoid)• TOA reflectance (right) normalization of sensor /solar geometry effects

Prod Acc% User Acc%Water 71.79 100Crop 79.17 79.17Trees 100 88.14Non-tree&bare 63.95 98.21Built 9.09 3.85Cloud&shadow 100 58.62Overall Acc = 79.7Kappa Coefficient = 0.73

Random Forests for classifying water in Rakhine near Ann:

Classification results of model with number of trees = 200, and max features = log2(n features)

Pixels incorrectly classified as water are scattered throughout the image in yellow, west of the identified river

Heavy growth and shadows from riparian vegetation hinder the model’s ability to identify some portions of small streams

• All individual Landsat scenes for years 2013-2017

Global Inland Water Surface Water Fraction

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

Average day and nighttime temperature of February 7 – 11, 2016

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

2013/12/192014/2/52014/2/212014/5/282014/10/32014/11/202015/1/232015/2/242015/10/222015/11/232016/2/112016/3/30

2013/4/232013/11/172014/2/52014/11/20

LCLU class Change components

surface water loss/gain (flooding*)

cropped areas flooding*built-up and impervioussurfaces flooding*

forest & forest patchflooding*, tree loss, vegetation condition (NDVI/EVI)

forest clearanceflooding*, tree gain, vegetation condition (NDVI/EVI)

short vegetation vegetation condition (NDVI/EVI)

bare and sparsely vegetated flooding*

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

Residential areas in Ann Township

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

P. falciparum

Source: pixnio.com Source: http://www.saburchill.com/ans02/chapters/chap013.html

Satellite data+

spatial data+

CRF surveys

Satellite data+

entomological surveys

usPCR testing

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

1 2 3 4 5

Malaria prevalence (all population)

P. falciparum P. vivax P. mixed (F&V)

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

1 2 3 4 5

Malaria prevalence (men)

PF prevalence total PV prevalence total

Pmixed (F&V)

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

1 2 3 4 5

Malaria prevalence (women)

PF prevalence total PV prevalence total

Pmixed (F&V)

0%

5%

10%

15%

20%

25%

30%

35%

40%

1 2 3 4 5

All malaria prevalence by age group (women)

2-10 years 11-17 years 18-40 years >40 years

0%

5%

10%

15%

20%

25%

30%

35%

40%

1 2 3 4 5

All malaria prevalence by age group (men)

2-10 years 11-17 years 18-40 years >40 years

Land cover/use baseline assessment (single time)

Human mobility pattern mapping from surveys and auxiliary data

Downscaled meteorological parameters (8-day)

Analyze malaria prevalence as a function of environmental conditions

Analyze malaria prevalence as a function of human mobility

Focus group interviews:Verify study findings in focus group interviews with health practitioners and researchers in the Rakhine state.

Dynamic mapping of surface conditions (8-day)

Identifying mobility networks

Objective 1: Land cover/use mapping

Objective 2: Mobility

Objective 3: Analytics Objective 4: Expert verification

Questions?