Malaria Vector in Northeastern Venezuela. Sarah Anne Guagliardo MPH candidate, 2010 Yale University School of Epidemiology and Public Health

Transcription

1 Vegetation associated with the An. Aquasalis Malaria Vector in Northeastern Venezuela Sarah Anne Guagliardo g MPH candidate, 2010 Yale University School of Epidemiology and Public Health

2 Outline Problem description and objectives Remote sensing background information Methodology Classification types Next steps

3 Problem Description High incidence of malaria in Cajigal Municipality in Sucre State, Venezuela Previous studies have shown that the Anopheles aquasalis principle vector for plasmodium vivax is found near mangroves (Grillet et al 2000) P li i l i di t th t th Preliminary analyses indicate that the mangroves are farther away than expected from towns with high malaria burden (Grillet, unpublished data)

4 Objectives Objectives : 1. Classify land cover and vegetation (specifically mangroves and swamps) in the areas around towns with high malaria burden using remote sensing 2. Describe the relationship between the presence of the vector and various types ofvegetation using logistic regression. 3. Develop a risk map incorporating epidemiological information with GIS.

5 What is remote sensing? Foto: Eastern Tennessee State University

6 Types of Sensors Passive sensors measure natural emitted and reflected electromagnetic energy. Active sensors Send pulses of electromagnetic energy (such as a laser or microwaves) and then observe the reflection of the energy. Measure emitted radiation from the satelite in addition to natural radiation.

7 Electromagnetic Spectrum Yale University Center for Earth Observation

8 Yale University Center for Earth Observation Layer Stack

9 Comparison of Spectral Ranges ASTER and Landsat Satellites Yale University Center for Earth Observation

10 Digital Numbers and Pixels Each pixel contains a digital number that represents the intensity of the emitted or reflected radiation for the pixel area. Each layer (or band ) in the image is capable of measuring a different spectral range of radiation. USGS

11 Yale University Center for Earth Observation Spectral Signatures

12 Examples of Satellites Useful for the Environmental Sciences AVHRR (NPOESS) GOES &MeteoSAT Landsat MODIS ASTER TRMM SPOT SeaWIFS QuikSCAT GRACE CloudSAT IceSAT IKONOS SAGE CALIPSO RadarSAT

13 Choosing an Image Spectral resolution Spatial resolution Temporal resolution Swath Cost Clouds

14 Study Area Sucre State Cajigal Municipality

15 Problem Description A germinans mangroves in Cajigal Municipality A. germinans mangroves in Cajigal Municipality Photo: Maria Eugenia Grillet

16 Population Size Rio Seco El Paujil

17 Elevation Elevation in meters < Towns Roads

18 Elevation Elevation in meters > An. aquasalis is found at elevations <550m above sea level Towns Roads

19 Annual Parasitic Index and Elevation Elevation in meters Annual Parasitic Index

20 Study Area Annual Parasitic Index = number of positive slides * 1000 total population

21 Image Acquisition and Processing Landsat TM ASTER December 24, 1990 February 15, 2002 Path h1, Row 53 Path 1, Row 53

22 Image Processing

23 Image Processing Image Subset

24 Image Processing Image subset with vector files Elevation in meters Annual Parasitic Index

25 Image Processing Area of interest

26 Image Classification Unsupervised classification Supervised classification

27 Unsupervised Classification Computer program automatically ti dt determines classes of pixels by examining the spectral signaturesof eachpixel pixel. User specifies the quantity of classes and other parameters ( such as standard deviation). Useful to see how the data are distributed, and to see what areas of the image might need further investigation.

28 Unsupervised Classifications with six nine classes Water /cloud shadows Mangroves Green areas Dark green areas Herbaceous swamps Towns/development Clouds

29 Supervised Classification The user selects training areas in the image. The program identifies other pixels similar to those in the training areas. Requires extensive knowledge of the study area.

30 Classification is an iterative process!

31 Ground Truthing

32 Ground Truthing Multiple supervised classifications were done using information from ground truthing, maps, Google Earth, and datacollected on previousfieldtrips Water Shallow water Mangroves Green areas Dark green areas Herbaceous swamps Towns/development Clouds Cloud shadows

33 Ground Truthing

34 Methodology Spectral signaturewas used to distinguishbetween mangroves and non mangrove vegetation. Mangrove electromagnetic reflection is lower in the near infrared region relative to other types of vegetation (Hirose et al 2004). Reflejo (m mm) Mangrove reflection is lower Cloud shadows Clouds Green areas Mangroves Nivel de Swamps Towns/ development Wt Water ASTER Bands 1 9

35 Methodology Using a subtraction formula, non mangrove pixels were masked from the image.

36 Methodology Unsupervised classification was used to determine differences among mangroves and to identify areas for further investigation.

37 Methodology More that 100 GPS points were collected in various regions around Cajigal jg Municipality. It is now necessary to incorporate this information to complete a new supervised classification.

38 1000m buffers around cloudless towns with high API Water Shallow water Mangroves Green areas Dark green areas Herbaceous swamps Towns/development Clouds Cloud shadows

39 Land cover for the Towns with the Highest APIs Water Shallow water Mangroves Green areas Dark green areas Herbaceous swamps Towns/development Clouds Cloud shadows

40 Land cover classification with 1km buffers around malarious areas. Water Shallow water Mangroves Green areas Dark green areas Herbaceous swamps Towns/development Clouds Cloud shadows

41 Tassled Cap Wetness Index Tassled Cap Wetness Index (Rainbow Reversed 99% Autoclip)

42 Five Classes Wetness Classification

43 Classified Wetness Index

44 Next Steps Complete another supervised classification of mangroves by species using the data collected over the summer. Check the validity of the classification by using finer spatial scale (ISPOT, IKONOS) UseGIS to experiment with various buffer sizes around populated areas and calculate land cover statistics. Use the Wetness Index to identify other areas of risk.

45 Limitations GPS points for towns do not line up with satellite images Image selection was based on % cloud coverage