Authors: Abolfazl Mollalo*, University of Florida
Topics: Medical and Health Geography, Environment, United States
Keywords: GIS, Zoonotic cutaneous leishmanias, Machine learning technique, Remote Sensing
Session Type: Paper
Start / End Time: 4:40 PM / 6:20 PM
Room: Lafayette, Marriott, River Tower Elevators, 41st Floor
Presentation File: No File Uploaded
Distribution and abundance of Phlebotomus papatasi, the main vector of zoonotic cutaneous leishmaniasis (ZCL), is one of the major public health threats in Iran. The main purpose of this paper is to model spatial distribution of this species and to identify risk-prone regions of the sand fly in an endemic area of cutaneous leishmaniasis to help decision makers for targeted interventions. We developed a geodatabase of the collected sandflies in 2014 from different parts of the study area. For every selected village, 30 indoor and 30 outdoor sticky traps coated with castor oil were used to collect sandflies. We also gathered and prepared associated environmental factors including topographic, weather parameters, distance to rivers and remotely sensed data (including normalized difference vegetation index (NDVI) and land surface temperature (LST)) in a GIS framework. Applicability of two important machine learning techniques namely random forest (RF) and support vector machine (SVM) for predicting spatial distribution of the vector was investigated and compared with logistic regression model. Predictive performances of the models were compared for both train and test (independent) datasets using area under ROC curve (AUC) and Kappa statistics. Sensitivity analysis performed to find the factors with the highest contribution in predicting presence of the species. Also, susceptible areas for vector presence were identified. Findings demonstrated that integration of machine learning techniques and environmental data in a GIS platform is a useful and cost-effective way in monitoring zoonotic cutaneous leishmaniasis.