NDVI and LST extraction of MODIS data under a GIS open source application - Rickettsia study case in Angola

Abstract

Fevers of unknown origin can have different aetiologies. The overlapping symptomatology of rickettsial infection and other endemic diseases that cause fever leads to a misdiagnosis or under-diagnosis of spotted fever group of Rickettsia (SFGR).

To better understand the epidemiology of this vector-borne disease in Angola, a comprehensive seroprevalence study was conducted investigating the exposure to SFGR in a sample of 92 febrile, Malaria and Yellow Fever negative human plasma specimen, collected to the study of the national surveillance of febrile syndromes between 2016 and 2017, in Angola.

The seroprevalence of IgG antibodies against SFG Rickettsia in humans was calculated by gender, and aimag (province). All data were analyzed through a logistic regression. Spatial data sources included Normalized Differential Vegetation Index (NDVI) and Land Surface Temperature (LST) products by Moderate Resolution Imaging Spectroradiometer (MODIS).

The main objective of this work was the development of a GIS open source application to automatize the extraction of LST and NDVI products from MODIS images. The application was created as a simple graphic interface composed by two input fields (the text file with the coordinates of the sampling points (in sinusoidal coordinate system and the folder with the MODIS images), the field to define the buffer distance, and the output file. The application was tested considering MOD11A1 (LST product), MOD13Q1 and MYD13Q1 (NDVI product), free download from the USGS.

QGIS 2.18.17 was used for geospatial operations and Python language was employed for the development of the GIS open source application under QGIS software. The process includes the circumscription of the major clusters where human data were collected. Then, a convex hull (minimum convex bounding geometry) was created around each sampling site with a 10 km buffer zone to accommodate the mobility among the nomadic people being samples. Counts of seropositive and seronegative humans were calculated within each of these sampling clusters along with the mean, maximum, and minimum values of NDVI and LST, and percent area of each land cover class.

The application was tested in a set of 92 points in Angola and a buffer of 10 km considering the Universal transverse Mercator (UTM) Zone 33S projection (EPSG:32733) was applied for each point. The LST and NDVI statistical values were extracted for each sampling cluster.

Variations in ecological niches, abundance of vegetation and land surface temperature, for ticks and fleas between different provinces could be in part responsible for the geographic differences in seroprevalence observed with SFGR.