Time series clustering applied to eco-epidemiology: the case of Aedes aegypti in Córdoba, Argentina.

VERONICA ANDREO; XIMENA PORCASI; CARLA RODRIGUEZ GONZALEZ; LAURA LOPEZ; CLAUDIO GUZMAN; MARCELO SCAVUZZO

Bahia Blanca

Congreso; IEEE RPIC 2019 - XVIII Reunión de trabajo en Procesamiento de la Información y Control; 2019

Instituto de Investigaciones en Ingeniería Eléctrica ?Alfredo Desages? (IIIE, Universidad Nacional del sur y CONICET) y el Departamento de Ingeniería Eléctrica y de Computadoras (DIEC, Universidad Nacional del Sur)

Dengue fever is one of the most widespread vector borne diseases in the world. The virus causing it is transmitted by Aedes aegypti mosquitoes. Prevention of outbreaks is mostly based on vector surveillance, removal of breeding sites and insecticide applications. However, mosquitoes activity varies in time and space. In this work we assessed the existence of groups of temporal patterns in time series of Aedes aegypti ovitraps records. Our hypothesis was that the identification of different temporal patterns might guide differential control interventions in the light of limited resources. Moreover, and with the ultimate goal of predicting risk in places with no mosquito field data, we associated the different temporal patterns with environmental variables derived from satellite imagery, assuming that differences in mosquito activity patterns in time are determined by local conditions of the environment. We tested  three different types of time series clustering algorithms. All of them identified three groups as the best splitting for the ovitraps time series data. Dynamic Time Warping (DTW) and Partition Around Medoids (PAM) algorithms showed the best separation according to internal cluster validity indices (CVI). They denoted certain consistency among them and also relationship with total egg counts. The groups depicting higher risk showed higher annual abundances and a peak between mid January and mid February. Although we did not find conclusive results in terms of local environmental factors related to these groups, they might  provide guidance on where and when to intensify surveillance, removal of breeding sites and insecticide spraying.